feat: cost tracking

feat: token comsumption usage
feat: hybrid compaction buffer (fixed tokens + ratio of context)
2026-04-23 15:34:07 -07:00 · 2026-04-23 15:05:30 -07:00 · 2026-04-23 15:04:19 -07:00 · 2026-04-22 21:38:21 -07:00 · 2026-04-22 21:33:33 -07:00 · 2026-04-22 21:27:24 -07:00
799 changed files with 73767 additions and 52578 deletions
@@ -1,4 +1,74 @@
 {
+  "permissions": {
+    "allow": [
+      "Bash(grep -n \"_is_context_too_large_error\" core/framework/agent_loop/agent_loop.py core/framework/agent_loop/internals/*.py)",
+      "Read(//^class/ {cls=$3} /def test_/**)",
+      "Read(//^    @pytest.mark.asyncio/{getline n; print NR\": \"n} /^    def test_/**)",
+      "Bash(python3)",
+      "Bash(grep -nE 'Tool\\\\\\(\\\\s*$|name=\"[a-z_]+\",' core/framework/tools/queen_lifecycle_tools.py)",
+      "Bash(awk -F'\"' '{print $2}')",
+      "Bash(grep -n \"create_colony\\\\|colony-spawn\\\\|colony_spawn\" /home/timothy/aden/hive/core/framework/agents/queen/nodes/__init__.py /home/timothy/aden/hive/core/framework/tools/*.py)",
+      "Bash(git stash:*)",
+      "Bash(python3 -c \"import sys,json; d=json.loads\\(sys.stdin.read\\(\\)\\); print\\('keys:', list\\(d.keys\\(\\)\\)[:10]\\)\")",
+      "Bash(python3 -c ':*)",
+      "Bash(uv run:*)",
+      "Read(//tmp/**)",
+      "Bash(grep -n \"useColony\\\\|const { queens, queenProfiles\" /home/timothy/aden/hive/core/frontend/src/pages/queen-dm.tsx)",
+      "Bash(awk 'NR==385,/\\\\}, \\\\[/' /home/timothy/aden/hive/core/frontend/src/pages/queen-dm.tsx)",
+      "Bash(xargs -I{} sh -c 'if ! grep -q \"^import base64\\\\|^from base64\" \"{}\"; then echo \"MISSING: {}\"; fi')",
+      "Bash(find /home/timothy/aden/hive/core/framework -name \"*.py\" -type f -exec grep -l \"FileConversationStore\\\\|class.*ConversationStore\" {} \\\\;)",
+      "Bash(find /home/timothy/aden/hive/core/framework -name \"*.py\" -exec grep -l \"run_parallel_workers\\\\|create_colony\" {} \\\\;)",
+      "Bash(awk '/^    async def execute\\\\\\(self, ctx: AgentContext\\\\\\)/,/^    async def [a-z_]+/ {print NR\": \"$0}' /home/timothy/aden/hive/core/framework/agent_loop/agent_loop.py)",
+      "Bash(grep -r \"max_concurrent_workers\\\\|max_depth\\\\|recursion\\\\|spawn.*bomb\" /home/timothy/aden/hive/core/framework/host/*.py)",
+      "Bash(wc -l /home/timothy/aden/hive/tools/src/gcu/browser/*.py /home/timothy/aden/hive/tools/src/gcu/browser/tools/*.py)",
+      "Bash(file /tmp/gcu_verify/*.png)",
+      "Bash(ps -eo pid,cmd)",
+      "Bash(ps -o pid,lstart,cmd -p 746640)",
+      "Bash(kill 746636)",
+      "Bash(ps -eo pid,lstart,cmd)",
+      "Bash(grep -E \"^d|\\\\.py$\")",
+      "Bash(grep -E \"\\\\.\\(ts|tsx\\)$\")",
+      "Bash(xargs cat:*)",
+      "Bash(find /home/timothy/aden/hive -path \"*/.venv\" -prune -o -name \"*.py\" -type f -exec grep -l \"frontend\\\\|UI\\\\|terminal\\\\|interactive\\\\|TUI\" {} \\\\;)",
+      "Bash(wc -l /home/timothy/.hive/backup/*/SKILL.md)",
+      "Bash(awk -F'::' '{print $1}')",
+      "Bash(wait)",
+      "Bash(pkill -f \"pytest.*test_event_loop_node\")",
+      "Bash(pkill -f \"pytest.*TestToolConcurrency\")",
+      "Bash(grep -n \"def.*discover\\\\|/api/agents\\\\|agents_discover\" /home/timothy/aden/hive/core/framework/server/*.py)",
+      "Bash(bun run:*)",
+      "Bash(npx eslint:*)",
+      "Bash(npm run:*)",
+      "Bash(npm test:*)",
+      "Bash(grep -n \"PIL\\\\|Image\\\\|to_thread\\\\|run_in_executor\" /home/timothy/aden/hive/tools/src/gcu/browser/*.py /home/timothy/aden/hive/tools/src/gcu/browser/tools/*.py)",
+      "WebFetch(domain:docs.litellm.ai)",
+      "Bash(cat /home/timothy/aden/hive/.venv/lib/python3.11/site-packages/litellm-*.dist-info/METADATA)",
+      "Bash(find \"/home/timothy/.hive/agents/queens/queen_brand_design/sessions/session_20260415_100751_d49f4c28/\" -type f -name \"*.json*\" -exec grep -l \"协日\" {} \\\\;)",
+      "Bash(grep -v ':0$')",
+      "Bash(curl -s -m 2 http://127.0.0.1:4002/sse -o /dev/null -w 'status=%{http_code} time=%{time_total}s\\\\n')",
+      "mcp__gcu-tools__browser_status",
+      "mcp__gcu-tools__browser_start",
+      "mcp__gcu-tools__browser_navigate",
+      "mcp__gcu-tools__browser_evaluate",
+      "mcp__gcu-tools__browser_screenshot",
+      "mcp__gcu-tools__browser_open",
+      "mcp__gcu-tools__browser_click_coordinate",
+      "mcp__gcu-tools__browser_get_rect",
+      "mcp__gcu-tools__browser_type_focused",
+      "mcp__gcu-tools__browser_wait",
+      "Bash(python3 -c ' *)",
+      "Bash(python3 scripts/debug_queen_prompt.py independent)",
+      "Bash(curl -s --max-time 2 http://127.0.0.1:9230/status)",
+      "Bash(python3 -c \"import json, sys; print\\(json.loads\\(sys.stdin.read\\(\\)\\)['data']['content']\\)\")",
+      "Bash(python3 -c \"import json; json.load\\(open\\('/home/timothy/aden/hive/tools/browser-extension/manifest.json'\\)\\)\")"
+    ],
+    "additionalDirectories": [
+      "/home/timothy/.hive/skills/writing-hive-skills",
+      "/tmp",
+      "/home/timothy/.hive/skills",
+      "/home/timothy/aden/hive/core/frontend/src/components"
+    ]
+  },
  "hooks": {
    "PostToolUse": [
      {
@@ -64,7 +64,7 @@ snapshot = await browser_snapshot(tab_id)
 |---------|--------------|-------|
 | Scroll doesn't move | Nested scroll container | Look for `overflow: scroll` divs |
 | Click no effect | Element covered | Check `getBoundingClientRect` vs viewport |
-| Type clears | Autocomplete/React | Check for event listeners on input |
+| Type clears | Autocomplete/React | Check for event listeners on input; try `browser_type_focused` |
 | Snapshot hangs | Huge DOM | Check node count in snapshot |
 | Snapshot stale | SPA hydration | Wait after navigation |

@@ -229,7 +229,7 @@ function queryShadow(selector) {
 |-------|-------------|----------|
 | Scroll not working | Find scrollable container | Mouse wheel at container center |
 | Click no effect | JavaScript click() | CDP mouse events |
-| Type clears | Add delay_ms | Use execCommand |
+| Type clears | Add delay_ms | Use `browser_type_focused` (Input.insertText) |
 | Snapshot hangs | Add timeout_s | DOM snapshot fallback |
 | Stale content | Wait for selector | Increase wait_until timeout |
 | Shadow DOM | Pierce selector | JavaScript traversal |
@@ -9,7 +9,6 @@ Fix: Add wait_for_selector between scroll calls

 import asyncio
 import sys
-import time
 from pathlib import Path

 sys.path.insert(0, str(Path(__file__).parent.parent.parent.parent / "tools" / "src"))
@@ -36,7 +35,7 @@ async def test_twitter_lazy_scroll():
            if bridge.is_connected:
                print("✓ Extension connected!")
                break
-            print(f"Waiting for extension... ({i+1}/10)")
+            print(f"Waiting for extension... ({i + 1}/10)")
        else:
            print("✗ Extension not connected")
            return
@@ -58,7 +57,7 @@ async def test_twitter_lazy_scroll():
        # Count initial tweets
        initial_count = await bridge.evaluate(
            tab_id,
-            '(function() { return document.querySelectorAll(\'[data-testid="tweet"]\').length; })()'
+            "(function() { return document.querySelectorAll('[data-testid=\"tweet\"]').length; })()",
        )
        print(f"Initial tweet count: {initial_count.get('result', 0)}")

@@ -70,7 +69,7 @@ async def test_twitter_lazy_scroll():
        print("\n--- Scrolling with waits ---")
        for i in range(3):
            result = await bridge.scroll(tab_id, "down", 500)
-            print(f"  Scroll {i+1}: {result.get('method', 'unknown')} method")
+            print(f"  Scroll {i + 1}: {result.get('method', 'unknown')} method")

            # Wait for new content to load
            await asyncio.sleep(2)
@@ -78,20 +77,20 @@ async def test_twitter_lazy_scroll():
            # Count tweets after scroll
            count_result = await bridge.evaluate(
                tab_id,
-                '(function() { return document.querySelectorAll(\'[data-testid="tweet"]\').length; })()'
+                "(function() { return document.querySelectorAll('[data-testid=\"tweet\"]').length; })()",
            )
-            count = count_result.get('result', 0)
+            count = count_result.get("result", 0)
            print(f"  Tweet count after scroll: {count}")

        # Final count
        final_count = await bridge.evaluate(
            tab_id,
-            '(function() { return document.querySelectorAll(\'[data-testid="tweet"]\').length; })()'
+            "(function() { return document.querySelectorAll('[data-testid=\"tweet\"]').length; })()",
        )
-        final = final_count.get('result', 0)
-        initial = initial_count.get('result', 0)
+        final = final_count.get("result", 0)
+        initial = initial_count.get("result", 0)

-        print(f"\n--- Results ---")
+        print("\n--- Results ---")
        print(f"Initial tweets: {initial}")
        print(f"Final tweets: {final}")

@@ -9,7 +9,6 @@ Fix: Find visible modal container (highest z-index scrollable), scroll that

 import asyncio
 import sys
-import time
 from pathlib import Path

 sys.path.insert(0, str(Path(__file__).parent.parent.parent.parent / "tools" / "src"))
@@ -60,7 +59,7 @@ async def test_modal_scroll():
        # Click button to open modal
        print("\n--- Opening modal ---")
        # Find and click the "Open Modal" button
-        result = await bridge.click(tab_id, '.ws-btn', timeout_ms=5000)
+        result = await bridge.click(tab_id, ".ws-btn", timeout_ms=5000)
        print(f"Click result: {result}")

        await asyncio.sleep(1)
@@ -52,7 +52,9 @@ async def test_overlay_click():
        <head><title>Overlay Test</title></head>
        <body>
            <button id="target-btn" onclick="alert('Clicked!')">Click Me</button>
-            <div id="overlay" style="position:fixed;top:0;left:0;width:100%;height:100%;background:rgba(0,0,0,0.3);z-index:1000;"></div>
+            <div id="overlay" style="position:fixed;top:0;left:0;
+            width:100%;height:100%;
+            background:rgba(0,0,0,0.3);z-index:1000;"></div>
            <script>
                window.clickCount = 0;
                document.getElementById('target-btn').addEventListener('click', () => {
@@ -65,6 +67,7 @@ async def test_overlay_click():

        # Navigate to data URL
        import base64
+
        data_url = f"data:text/html;base64,{base64.b64encode(test_html.encode()).decode()}"
        await bridge.navigate(tab_id, data_url, wait_until="load")

@@ -91,7 +94,7 @@ async def test_overlay_click():
                    targetElement: btn.tagName
                };
            })();
-        """
+        """,
        )
        print(f"Coverage check: {coverage_check.get('result', {})}")

@@ -100,10 +103,7 @@ async def test_overlay_click():
        print(f"Click result: {click_result}")

        # Check if click registered
-        count_result = await bridge.evaluate(
-            tab_id,
-            "(function() { return window.clickCount; })()"
-        )
+        count_result = await bridge.evaluate(tab_id, "(function() { return window.clickCount; })()")
        count = count_result.get("result", 0)
        print(f"Click count after CDP click: {count}")

@@ -10,7 +10,6 @@ Fix: Use piercing selector (host >>> target) or traverse shadow roots

 import asyncio
 import sys
-import base64
 from pathlib import Path

 sys.path.insert(0, str(Path(__file__).parent.parent.parent.parent / "tools" / "src"))
@@ -100,7 +99,7 @@ async def test_shadow_dom():
                });
                return { count: hosts.length, hosts };
            })();
-        """
+        """,
        )
        print(f"Shadow DOM detection: {detection.get('result', {})}")

@@ -126,15 +125,12 @@ async def test_shadow_dom():
                }
                return { success: false, error: 'Button not found' };
            })();
-        """
+        """,
        )
        print(f"JS click result: {click_result.get('result', {})}")

        # Verify click was registered
-        count_result = await bridge.evaluate(
-            tab_id,
-            "(function() { return window.shadowClickCount || 0; })()"
-        )
+        count_result = await bridge.evaluate(tab_id, "(function() { return window.shadowClickCount || 0; })()")
        count = count_result.get("result") or 0
        print(f"Shadow click count: {count}")

@@ -10,7 +10,6 @@ Fix: Focus via JavaScript, use execCommand('insertText') or Input.dispatchKeyEve

 import asyncio
 import sys
-import base64
 from pathlib import Path

 sys.path.insert(0, str(Path(__file__).parent.parent.parent.parent / "tools" / "src"))
@@ -54,10 +53,14 @@ async def test_contenteditable():
            <h2>ContentEditable Test</h2>

            <h3>1. Simple contenteditable div</h3>
-            <div id="editor1" contenteditable="true" style="border:1px solid #ccc;padding:10px;min-height:50px;">Start text</div>
+            <div id="editor1" contenteditable="true"
+            style="border:1px solid #ccc;padding:10px;
+            min-height:50px;">Start text</div>

            <h3>2. Rich text editor (like Notion)</h3>
-            <div id="editor2" contenteditable="true" style="border:1px solid #ccc;padding:10px;min-height:50px;">
+            <div id="editor2" contenteditable="true"
+            style="border:1px solid #ccc;padding:10px;
+            min-height:50px;">
                <p>Type here...</p>
            </div>

@@ -106,7 +109,7 @@ async def test_contenteditable():
                    ids: Array.from(editables).map(el => el.id)
                };
            })();
-        """
+        """,
        )
        print(f"Contenteditable detection: {detection.get('result', {})}")

@@ -115,8 +118,7 @@ async def test_contenteditable():
        await bridge.click(tab_id, "#input1")
        await bridge.type_text(tab_id, "#input1", "Hello input")
        input_result = await bridge.evaluate(
-            tab_id,
-            "(function() { return document.getElementById('input1').value; })()"
+            tab_id, "(function() { return document.getElementById('input1').value; })()"
        )
        print(f"Input value: {input_result.get('result', '')}")

@@ -126,7 +128,7 @@ async def test_contenteditable():
        await bridge.type_text(tab_id, "#editor1", "Hello contenteditable", clear_first=True)
        editor_result = await bridge.evaluate(
            tab_id,
-            "(function() { return document.getElementById('editor1').innerText; })()"
+            "(function() { return document.getElementById('editor1').innerText; })()",
        )
        print(f"Editor1 innerText: {editor_result.get('result', '')}")

@@ -142,7 +144,7 @@ async def test_contenteditable():
                document.execCommand('insertText', false, 'Hello from execCommand');
                return editor.innerText;
            })();
-        """
+        """,
        )
        print(f"Editor2 after execCommand: {insert_result.get('result', '')}")

@@ -10,7 +10,6 @@ Fix: Add delay_ms between keystrokes

 import asyncio
 import sys
-import base64
 from pathlib import Path

 sys.path.insert(0, str(Path(__file__).parent.parent.parent.parent / "tools" / "src"))
@@ -87,7 +86,26 @@ async def test_autocomplete():
            <div id="log" style="margin-top:20px;font-family:monospace;"></div>

            <script>
-                const countries = ["Afghanistan","Albania","Algeria","Andorra","Angola","Argentina","Armenia","Australia","Austria","Azerbaijan","Bahamas","Bahrain","Bangladesh","Belarus","Belgium","Belize","Benin","Bhutan","Bolivia","Brazil","Canada","China","Colombia","Denmark","Egypt","France","Germany","India","Indonesia","Italy","Japan","Mexico","Netherlands","Nigeria","Norway","Pakistan","Peru","Philippines","Poland","Portugal","Russia","Spain","Sweden","Switzerland","Thailand","Turkey","Ukraine","United Kingdom","United States","Vietnam"];
+                const countries = [
+                    "Afghanistan","Albania","Algeria",
+                    "Andorra","Angola","Argentina",
+                    "Armenia","Australia","Austria",
+                    "Azerbaijan","Bahamas","Bahrain",
+                    "Bangladesh","Belarus","Belgium",
+                    "Belize","Benin","Bhutan",
+                    "Bolivia","Brazil","Canada",
+                    "China","Colombia","Denmark",
+                    "Egypt","France","Germany",
+                    "India","Indonesia","Italy",
+                    "Japan","Mexico","Netherlands",
+                    "Nigeria","Norway","Pakistan",
+                    "Peru","Philippines","Poland",
+                    "Portugal","Russia","Spain",
+                    "Sweden","Switzerland","Thailand",
+                    "Turkey","Ukraine",
+                    "United Kingdom","United States",
+                    "Vietnam"
+                ];

                const input = document.getElementById('search');
                const log = document.getElementById('log');
@@ -126,11 +144,15 @@ async def test_autocomplete():
                        div.setAttribute('class', 'autocomplete-items');
                        this.parentNode.appendChild(div);

-                        countries.filter(c => c.substr(0, val.length).toUpperCase() === val.toUpperCase())
-                            .slice(0, 5)
-                            .forEach(country => {
+                        countries.filter(
+                            c => c.substr(0, val.length).toUpperCase()
+                                === val.toUpperCase()
+                        ).slice(0, 5).forEach(country => {
                                const item = document.createElement('div');
-                                item.innerHTML = '<strong>' + country.substr(0, val.length) + '</strong>' + country.substr(val.length);
+                                item.innerHTML = '<strong>'
+                                    + country.substr(0, val.length)
+                                    + '</strong>'
+                                    + country.substr(val.length);
                                item.addEventListener('click', function() {
                                    input.value = country;
                                    closeAllLists();
@@ -172,17 +194,13 @@ async def test_autocomplete():
        await asyncio.sleep(0.5)

        fast_result = await bridge.evaluate(
-            tab_id,
-            "(function() { return document.getElementById('search').value; })()"
+            tab_id, "(function() { return document.getElementById('search').value; })()"
        )
        fast_value = fast_result.get("result", "")
        print(f"Value after fast typing: '{fast_value}'")

        # Check events
-        events_result = await bridge.evaluate(
-            tab_id,
-            "(function() { return window.inputEvents; })()"
-        )
+        events_result = await bridge.evaluate(tab_id, "(function() { return window.inputEvents; })()")
        print(f"Events logged: {events_result.get('result', [])}")

        # Test 2: Slow typing (with delay) - should work
@@ -192,8 +210,7 @@ async def test_autocomplete():
        await asyncio.sleep(0.5)

        slow_result = await bridge.evaluate(
-            tab_id,
-            "(function() { return document.getElementById('search').value; })()"
+            tab_id, "(function() { return document.getElementById('search').value; })()"
        )
        slow_value = slow_result.get("result", "")
        print(f"Value after slow typing: '{slow_value}'")
@@ -201,7 +218,7 @@ async def test_autocomplete():
        # Check if dropdown appeared
        dropdown_result = await bridge.evaluate(
            tab_id,
-            "(function() { return document.querySelectorAll('.autocomplete-items div').length; })()"
+            "(function() { return document.querySelectorAll('.autocomplete-items div').length; })()",
        )
        dropdown_count = dropdown_result.get("result", 0)
        print(f"Dropdown items: {dropdown_count}")
@@ -87,10 +87,7 @@ async def test_huge_dom():
        await bridge.navigate(tab_id, data_url, wait_until="load")

        # Count elements
-        count_result = await bridge.evaluate(
-            tab_id,
-            "(function() { return document.querySelectorAll('*').length; })()"
-        )
+        count_result = await bridge.evaluate(tab_id, "(function() { return document.querySelectorAll('*').length; })()")
        elem_count = count_result.get("result", 0)
        print(f"DOM elements: {elem_count}")

@@ -126,10 +123,7 @@ async def test_huge_dom():
        await bridge.navigate(tab_id, "https://www.linkedin.com/feed", wait_until="load", timeout_ms=30000)
        await asyncio.sleep(2)

-        count_result = await bridge.evaluate(
-            tab_id,
-            "(function() { return document.querySelectorAll('*').length; })()"
-        )
+        count_result = await bridge.evaluate(tab_id, "(function() { return document.querySelectorAll('*').length; })()")
        elem_count = count_result.get("result", 0)
        print(f"LinkedIn DOM elements: {elem_count}")

@@ -11,7 +11,6 @@ Fix: Use wait_until="networkidle" or wait_for_selector
 import asyncio
 import sys
 import time
-import base64
 from pathlib import Path

 sys.path.insert(0, str(Path(__file__).parent.parent.parent.parent / "tools" / "src"))
@@ -82,9 +81,13 @@ async def test_spa_navigation():

                    // Render content
                    const content = {
-                        home: '<h1>Home Page</h1><p>Welcome to the SPA!</p><button id="home-btn">Home Action</button>',
-                        about: '<h1>About Page</h1><p>This is a simulated SPA.</p><button id="about-btn">About Action</button>',
-                        contact: '<h1>Contact Page</h1><p>Contact us at test@example.com</p><button id="contact-btn">Contact Action</button>'
+                        home: '<h1>Home Page</h1><p>Welcome!</p>'
+                            + '<button id="home-btn">Home Action</button>',
+                        about: '<h1>About Page</h1><p>Simulated SPA.</p>'
+                            + '<button id="about-btn">About Action</button>',
+                        contact: '<h1>Contact Page</h1>'
+                            + '<p>Contact us at test@example.com</p>'
+                            + '<button id="contact-btn">Contact Action</button>'
                    };

                    document.getElementById('app').innerHTML = content[page] || '<h1>404</h1>';
@@ -121,7 +124,7 @@ async def test_spa_navigation():
        # Check content immediately
        content = await bridge.evaluate(
            tab_id,
-            "(function() { return document.getElementById('app').innerText; })()"
+            "(function() { return document.getElementById('app').innerText; })()",
        )
        print(f"Content immediately after load: '{content.get('result', '')}'")

@@ -137,7 +140,7 @@ async def test_spa_navigation():
        # Check content after wait
        content_after = await bridge.evaluate(
            tab_id,
-            "(function() { return document.getElementById('app').innerText; })()"
+            "(function() { return document.getElementById('app').innerText; })()",
        )
        print(f"Content after wait: '{content_after.get('result', '')}'")

@@ -151,7 +154,7 @@ async def test_spa_navigation():
        # Check if content changed
        about_content = await bridge.evaluate(
            tab_id,
-            "(function() { return document.getElementById('app').innerText; })()"
+            "(function() { return document.getElementById('app').innerText; })()",
        )
        print(f"Content after SPA nav: '{about_content.get('result', '')}'")

@@ -167,7 +170,7 @@ async def test_spa_navigation():
        # Check content immediately
        content_networkidle = await bridge.evaluate(
            tab_id,
-            "(function() { return document.getElementById('app').innerText; })()"
+            "(function() { return document.getElementById('app').innerText; })()",
        )
        print(f"Content after networkidle: '{content_networkidle.get('result', '')}'")

@@ -44,7 +44,7 @@ def check_png(data: str) -> bool:
    """Verify that base64 data decodes to a valid PNG."""
    try:
        raw = base64.b64decode(data)
-        return raw[:8] == b'\x89PNG\r\n\x1a\n'
+        return raw[:8] == b"\x89PNG\r\n\x1a\n"
    except Exception:
        return False

@@ -218,7 +218,7 @@ async def main():
            if bridge.is_connected:
                print("✓ Extension connected!")
                break
-            print(f"Waiting for extension... ({i+1}/10)")
+            print(f"Waiting for extension... ({i + 1}/10)")
        else:
            print("✗ Extension not connected. Ensure Chrome with Beeline extension is running.")
            return
@@ -97,7 +97,7 @@ async def test_problematic_site(bridge: BeelineBridge, tab_id: int) -> dict:
                });
                return results;
            })();
-        """
+        """,
        )
        print(f"  Before scroll: {before.get('result', {})}")

@@ -126,7 +126,7 @@ async def test_problematic_site(bridge: BeelineBridge, tab_id: int) -> dict:
                });
                return results;
            })();
-        """
+        """,
        )
        print(f"  After scroll: {after.get('result', {})}")

@@ -194,7 +194,7 @@ async def detect_root_cause(bridge: BeelineBridge, tab_id: int) -> dict:
                largest: candidates[0]
            };
        })();
-    """
+    """,
    )
    detections["nested_scroll"] = scroll_check.get("result", {})
    print(f"  Nested scroll containers: {detections['nested_scroll']}")
@@ -212,7 +212,7 @@ async def detect_root_cause(bridge: BeelineBridge, tab_id: int) -> dict:
            });
            return { count: withShadow.length, elements: withShadow.slice(0, 5) };
        })();
-    """
+    """,
    )
    detections["shadow_dom"] = shadow_check.get("result", {})
    print(f"  Shadow DOM: {detections['shadow_dom']}")
@@ -225,7 +225,7 @@ async def detect_root_cause(bridge: BeelineBridge, tab_id: int) -> dict:
            const iframes = document.querySelectorAll('iframe');
            return { count: iframes.length };
        })();
-    """
+    """,
    )
    detections["iframes"] = iframe_check.get("result", {})
    print(f"  iframes: {detections['iframes']}")
@@ -240,7 +240,7 @@ async def detect_root_cause(bridge: BeelineBridge, tab_id: int) -> dict:
                body_children: document.body.children.length
            };
        })();
-    """
+    """,
    )
    detections["dom_size"] = dom_check.get("result", {})
    print(f"  DOM size: {detections['dom_size']}")
@@ -256,7 +256,7 @@ async def detect_root_cause(bridge: BeelineBridge, tab_id: int) -> dict:
                angular: !!document.querySelector('[ng-app], [ng-version]')
            };
        })();
-    """
+    """,
    )
    detections["frameworks"] = framework_check.get("result", {})
    print(f"  Frameworks: {detections['frameworks']}")
@@ -289,7 +289,7 @@ async def main():
            if bridge.is_connected:
                print("✓ Extension connected!")
                break
-            print(f"Waiting for extension... ({i+1}/10)")
+            print(f"Waiting for extension... ({i + 1}/10)")
        else:
            print("✗ Extension not connected. Ensure Chrome with Beeline extension is running.")
            return
@@ -1,18 +0,0 @@
-This project uses ruff for Python linting and formatting.
-
-Rules:
- Line length: 100 characters
- Python target: 3.11+
- Use double quotes for strings
- Sort imports with isort (ruff I rules): stdlib, third-party, first-party (framework), local
- Combine as-imports
- Use type hints on all function signatures
- Use `from __future__ import annotations` for modern type syntax
- Raise exceptions with `from` in except blocks (B904)
- No unused imports (F401), no unused variables (F841)
- Prefer list/dict/set comprehensions over map/filter (C4)
-
-Run `make lint` to auto-fix, `make check` to verify without modifying files.
-Run `make format` to apply ruff formatting.
-
-The ruff config lives in core/pyproject.toml under [tool.ruff].
@@ -1,35 +0,0 @@
-# Git
-.git/
-.gitignore
-
-# Documentation
-*.md
-docs/
-LICENSE
-
-# IDE
-.idea/
-.vscode/
-
-# Dependencies (rebuilt in container)
-node_modules/
-
-# Build artifacts
-dist/
-build/
-coverage/
-
-# Environment files
-.env*
-config.yaml
-
-# Logs
-*.log
-logs/
-
-# OS
-.DS_Store
-Thumbs.db
-
-# GitHub
-.github/
@@ -22,3 +22,6 @@ indent_size = 2

 [Makefile]
 indent_style = tab
+
+[*.{sh,ps1}]
+end_of_line = lf
@@ -16,7 +16,6 @@

 # Shell scripts (must use LF)
 *.sh text eol=lf
-quickstart.sh text eol=lf

 # PowerShell scripts (Windows-friendly)
 *.ps1 text eol=lf
@@ -122,3 +121,8 @@ CODE_OF_CONDUCT* text
 *.db binary
 *.sqlite binary
 *.sqlite3 binary
+
+# Lockfiles — mark generated so GitHub collapses them in PR diffs
+*.lock            linguist-generated=true -diff
+package-lock.json linguist-generated=true -diff
+uv.lock           linguist-generated=true -diff
@@ -63,7 +63,7 @@ jobs:
        working-directory: core
        run: |
          uv sync
-          uv run pytest tests/ -v
+          uv run pytest tests/ -v --ignore=tests/dummy_agents

  test-tools:
    name: Test Tools (${{ matrix.os }})
@@ -70,6 +70,8 @@ tmp/
 temp/

 exports/*
+exports.old*
+artifacts/*

 .claude/settings.local.json

@@ -79,3 +81,4 @@ core/tests/*dumps/*
 screenshots/*

 .gemini/*
+.coverage
@@ -0,0 +1,9 @@
+{"type": "connection", "event": "connect", "ts": "2026-04-04T01:10:38.245667+00:00", "profile": "default"}
+{"type": "connection", "event": "hello", "details": {"version": "1.0"}, "ts": "2026-04-04T01:10:38.247207+00:00", "profile": "default"}
+{"type": "connection", "event": "disconnect", "ts": "2026-04-04T01:11:57.148273+00:00", "profile": "default"}
+{"type": "connection", "event": "connect", "ts": "2026-04-04T01:12:09.162378+00:00", "profile": "default"}
+{"type": "connection", "event": "hello", "details": {"version": "1.0"}, "ts": "2026-04-04T01:12:09.163899+00:00", "profile": "default"}
+{"type": "connection", "event": "disconnect", "ts": "2026-04-04T01:15:12.826042+00:00", "profile": "default"}
+{"type": "connection", "event": "connect", "ts": "2026-04-04T01:15:30.842533+00:00", "profile": "default"}
+{"type": "connection", "event": "hello", "details": {"version": "1.0"}, "ts": "2026-04-04T01:15:30.845025+00:00", "profile": "default"}
+{"type": "tool_call", "tool": "browser_stop", "params": {"profile": "gcu-browser-worker:3"}, "result": {"ok": true, "status": "not_running", "profile": "gcu-browser-worker:3"}, "ok": true, "duration_ms": 0.01, "ts": "2026-04-04T01:29:04.294954+00:00", "profile": "default"}
@@ -1,3 +0,0 @@
-{
-  "mcpServers": {}
-}
@@ -333,6 +333,22 @@ make test-live     # Run live API integration tests (requires credentials)
 - **WebSocket** for real-time updates
 - **Tailwind CSS** for styling

+### Frontend Dev Workflow
+
+> **Note:** `./quickstart.sh` handles the full setup including the web UI.
+> The commands below are for contributors iterating on the frontend code after
+> initial setup is complete.
+
+```bash
+# Start the backend server
+hive serve
+
+# In a separate terminal, run the frontend dev server with hot-reload
+cd core/frontend
+npm install   # only needed after dependency changes
+npm run dev
+```
+
 ### Useful Development Commands

 ```bash
@@ -943,7 +959,7 @@ uv run pytest -m "not live"
 **Unit Test**
 ```python
 import pytest
-from framework.graph.node import Node
+from framework.orchestrator import NodeSpec as Node

 def test_node_creation():
    node = Node(id="test", name="Test Node", node_type="event_loop")
@@ -961,8 +977,8 @@ async def test_node_execution():
 **Integration Test**
 ```python
 import pytest
-from framework.graph.executor import GraphExecutor
-from framework.graph.node import Node
+from framework.orchestrator.orchestrator import Orchestrator as GraphExecutor
+from framework.orchestrator import NodeSpec as Node

@pytest.mark.asyncio
 async def test_graph_execution_with_multiple_nodes():
@@ -28,7 +28,7 @@ check: ## Run all checks without modifying files (CI-safe)
 	cd tools && uv run ruff format --check .

 test: ## Run all tests (core + tools, excludes live)
-	cd core && uv run python -m pytest tests/ -v
+	cd core && uv run python -m pytest tests/ -v --ignore=tests/dummy_agents
 	cd tools && uv run python -m pytest -v

 test-tools: ## Run tool tests only (mocked, no credentials needed)
@@ -38,7 +38,7 @@ test-live: ## Run live integration tests (requires real API credentials)
 	cd tools && uv run python -m pytest -m live -s -o "addopts=" --log-cli-level=INFO

 test-all: ## Run everything including live tests
-	cd core && uv run python -m pytest tests/ -v
+	cd core && uv run python -m pytest tests/ -v --ignore=tests/dummy_agents
 	cd tools && uv run python -m pytest -v
 	cd tools && uv run python -m pytest -m live -s -o "addopts=" --log-cli-level=INFO

@@ -1,5 +1,5 @@
 <p align="center">
-  <img width="100%" alt="Hive Banner" src="https://github.com/user-attachments/assets/a027429b-5d3c-4d34-88e4-0feaeaabbab3" />
+  <img width="100%" alt="Hive Banner" src="https://asset.acho.io/github/img/banner.gif" />
 </p>

 <p align="center">
@@ -40,7 +40,16 @@

 ## Overview

-Hive is a runtime harness for AI agents in production. You describe your goal in natural language; a coding agent (the queen) generates the agent graph and connection code to achieve it. During execution, the harness manages state isolation, checkpoint-based crash recovery, cost enforcement, and real-time observability. When agents fail, the framework captures failure data, evolves the graph through the coding agent, and redeploys automatically. Built-in human-in-the-loop nodes, browser control, credential management, and parallel execution give you production reliability without sacrificing adaptability.
+OpenHive is a zero-setup, model-agnostic execution harness that dynamically generates multi-agent topologies to tackle complex, long-running business workflows without requiring any orchestration boilerplate. By simply defining your objective, the runtime compiles a strict, graph-based execution DAG that safely coordinates specialized agents to execute concurrent tasks in parallel. Backed by persistent, role-based memory that intelligently evolves with your project's context, OpenHive ensures deterministic fault tolerance, deep state observability, and seamless asynchronous execution across whichever underlying LLMs you choose to plug in.
+
+## Features
+
+- ✅ Multi-Agent Coordination for parallel task execution 
+- ✅ Graph-based execution for recurring and complex processes 
+- ✅ Role-based memory that evolves with your projects 
+- ✅ Zero Setup - No technical configuration required
+- ✅ General Compute Use and Browser Use with Native Extension 
+- ✅ Custom Model Support

 Visit [adenhq.com](https://adenhq.com) for complete documentation, examples, and guides.

@@ -51,7 +60,7 @@ https://github.com/user-attachments/assets/bf10edc3-06ba-48b6-98ba-d069b15fb69d

 ## Who Is Hive For?

-Hive is the harness layer for teams moving AI agents from prototype to production. Models are getting better on their own — the bottleneck is the infrastructure around them: state management, failure recovery, cost control, and observability.
+Hive is the multi-agent harness layer for teams moving AI agents from prototype to production. Single agents like Openclaw and Cowork can finish personal jobs pretty well but lack the rigor to fulfil business processes. 

 Hive is a good fit if you:

@@ -139,17 +148,6 @@ Now you can run an agent by selecting the agent (either an existing agent or exa

 <img width="2549" height="1174" alt="Screenshot 2026-03-12 at 9 27 36 PM" src="https://github.com/user-attachments/assets/7c7d30fa-9ceb-4c23-95af-b1caa405547d" />

-## Features
-
- **Browser-Use** - Control the browser on your computer to achieve hard tasks
- **Parallel Execution** - Execute the generated graph in parallel. This way you can have multiple agents completing the jobs for you
- **[Goal-Driven Generation](docs/key_concepts/goals_outcome.md)** - Define objectives in natural language; the coding agent generates the agent graph and connection code to achieve them
- **[Adaptiveness](docs/key_concepts/evolution.md)** - Framework captures failures, calibrates according to the objectives, and evolves the agent graph
- **[Dynamic Node Connections](docs/key_concepts/graph.md)** - No predefined edges; connection code is generated by any capable LLM based on your goals
- **SDK-Wrapped Nodes** - Every node gets a shared data buffer, local RLM memory, monitoring, tools, and LLM access out of the box
- **[Human-in-the-Loop](docs/key_concepts/graph.md#human-in-the-loop)** - Intervention nodes that pause execution for human input with configurable timeouts and escalation
- **Real-time Observability** - WebSocket streaming for live monitoring of agent execution, decisions, and node-to-node communication
-
 ## Integration

 <a href="https://github.com/aden-hive/hive/tree/main/tools/src/aden_tools/tools"><img width="100%" alt="Integration" src="https://github.com/user-attachments/assets/a1573f93-cf02-4bb8-b3d5-b305b05b1e51" /></a>
@@ -194,18 +192,6 @@ flowchart LR
    style V6 fill:#fff,stroke:#ed8c00,stroke-width:1px,color:#cc5d00
 ```

-### The Hive Advantage
-
-| Typical Agent Frameworks   | Hive                                   |
-| -------------------------- | -------------------------------------- |
-| Focus on model orchestration | **Production harness**: state, recovery, observability |
-| Hardcode agent workflows   | Describe goals in natural language     |
-| Manual graph definition    | Auto-generated agent graphs            |
-| Reactive error handling    | Outcome-evaluation and adaptiveness    |
-| Static tool configurations | Dynamic SDK-wrapped nodes              |
-| Separate monitoring setup  | Built-in real-time observability       |
-| DIY budget management      | Integrated cost controls & degradation |
-
 ### How It Works

 1. **[Define Your Goal](docs/key_concepts/goals_outcome.md)** → Describe what you want to achieve in plain English
@@ -221,131 +207,6 @@ flowchart LR
 - [Configuration Guide](docs/configuration.md) - All configuration options
 - [Architecture Overview](docs/architecture/README.md) - System design and structure

-## Roadmap
-
-Aden Hive Agent Framework aims to help developers build outcome-oriented, self-adaptive agents. See [roadmap.md](docs/roadmap.md) for details.
-
-```mermaid
-flowchart TB
-    %% Main Entity
-    User([User])
-
-    %% =========================================
-    %% EXTERNAL EVENT SOURCES
-    %% =========================================
-    subgraph ExtEventSource [External Event Source]
-        E_Sch["Schedulers"]
-        E_WH["Webhook"]
-        E_SSE["SSE"]
-    end
-
-    %% =========================================
-    %% SYSTEM NODES
-    %% =========================================
-    subgraph WorkerBees [Worker Bees]
-        WB_C["Conversation"]
-        WB_SP["System prompt"]
-
-        subgraph Graph [Graph]
-            direction TB
-            N1["Node"] --> N2["Node"] --> N3["Node"]
-            N1 -.-> AN["Active Node"]
-            N2 -.-> AN
-            N3 -.-> AN
-
-            %% Nested Event Loop Node
-            subgraph EventLoopNode [Event Loop Node]
-                ELN_L["listener"]
-                ELN_SP["System Prompt<br/>(Task)"]
-                ELN_EL["Event loop"]
-                ELN_C["Conversation"]
-            end
-        end
-    end
-
-    subgraph JudgeNode [Judge]
-        J_C["Criteria"]
-        J_P["Principles"]
-        J_EL["Event loop"] <--> J_S["Scheduler"]
-    end
-
-    subgraph QueenBee [Queen Bee]
-        QB_SP["System prompt"]
-        QB_EL["Event loop"]
-        QB_C["Conversation"]
-    end
-
-    subgraph Infra [Infra]
-        SA["Sub Agent"]
-        TR["Tool Registry"]
-        WTM["Write through Conversation Memory<br/>(Logs/RAM/Harddrive)"]
-        SM["Shared Memory<br/>(State/Harddrive)"]
-        EB["Event Bus<br/>(RAM)"]
-        CS["Credential Store<br/>(Harddrive/Cloud)"]
-    end
-
-    subgraph PC [PC]
-        B["Browser"]
-        CB["Codebase<br/>v 0.0.x ... v n.n.n"]
-    end
-
-    %% =========================================
-    %% CONNECTIONS & DATA FLOW
-    %% =========================================
-
-    %% External Event Routing
-    E_Sch --> ELN_L
-    E_WH --> ELN_L
-    E_SSE --> ELN_L
-    ELN_L -->|"triggers"| ELN_EL
-
-    %% User Interactions
-    User -->|"Talk"| WB_C
-    User -->|"Talk"| QB_C
-    User -->|"Read/Write Access"| CS
-
-    %% Inter-System Logic
-    ELN_C <-->|"Mirror"| WB_C
-    WB_C -->|"Focus"| AN
-
-    WorkerBees -->|"Inquire"| JudgeNode
-    JudgeNode -->|"Approve"| WorkerBees
-
-    %% Judge Alignments
-    J_C <-.->|"aligns"| WB_SP
-    J_P <-.->|"aligns"| QB_SP
-
-    %% Escalate path
-    J_EL -->|"Report (Escalate)"| QB_EL
-
-    %% Pub/Sub Logic
-    AN -->|"publish"| EB
-    EB -->|"subscribe"| QB_C
-
-    %% Infra and Process Spawning
-    ELN_EL -->|"Spawn"| SA
-    SA -->|"Inform"| ELN_EL
-    SA -->|"Starts"| B
-    B -->|"Report"| ELN_EL
-    TR -->|"Assigned"| ELN_EL
-    CB -->|"Modify Worker Bee"| WB_C
-
-    %% =========================================
-    %% SHARED MEMORY & LOGS ACCESS
-    %% =========================================
-
-    %% Worker Bees Access (link to node inside Graph subgraph)
-    AN <-->|"Read/Write"| WTM
-    AN <-->|"Read/Write"| SM
-
-    %% Queen Bee Access
-    QB_C <-->|"Read/Write"| WTM
-    QB_EL <-->|"Read/Write"| SM
-
-    %% Credentials Access
-    CS -->|"Read Access"| QB_C
-```
-
 ## Contributing
 We welcome contributions from the community! We’re especially looking for help building tools, integrations, and example agents for the framework ([check #2805](https://github.com/aden-hive/hive/issues/2805)). If you’re interested in extending its functionality, this is the perfect place to start. Please see [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.

@@ -52,9 +52,7 @@ _DEFAULT_REDIRECT_PORT = 51121
 # This project reverse-engineered and published the public OAuth credentials
 # for Google's Antigravity/Cloud Code Assist API.
 # Source: https://github.com/NoeFabris/opencode-antigravity-auth
-_CREDENTIALS_URL = (
-    "https://raw.githubusercontent.com/NoeFabris/opencode-antigravity-auth/dev/src/constants.ts"
-)
+_CREDENTIALS_URL = "https://raw.githubusercontent.com/NoeFabris/opencode-antigravity-auth/dev/src/constants.ts"

 # Cached credentials fetched from public source
 _cached_client_id: str | None = None
@@ -68,9 +66,7 @@ def _fetch_credentials_from_public_source() -> tuple[str | None, str | None]:
        return _cached_client_id, _cached_client_secret

    try:
-        req = urllib.request.Request(
-            _CREDENTIALS_URL, headers={"User-Agent": "Hive-Antigravity-Auth/1.0"}
-        )
+        req = urllib.request.Request(_CREDENTIALS_URL, headers={"User-Agent": "Hive-Antigravity-Auth/1.0"})
        with urllib.request.urlopen(req, timeout=10) as resp:
            content = resp.read().decode("utf-8")
            import re
@@ -168,10 +164,7 @@ class OAuthCallbackHandler(BaseHTTPRequestHandler):
            if "code" in query and "state" in query:
                OAuthCallbackHandler.auth_code = query["code"][0]
                OAuthCallbackHandler.state = query["state"][0]
-                self._send_response(
-                    "Authentication successful! You can close this window "
-                    "and return to the terminal."
-                )
+                self._send_response("Authentication successful! You can close this window and return to the terminal.")
                return

        self._send_response("Waiting for authentication...")
@@ -296,8 +289,7 @@ def validate_credentials(access_token: str, project_id: str = _DEFAULT_PROJECT_I
        "Authorization": f"Bearer {access_token}",
        "Content-Type": "application/json",
        "User-Agent": (
-            "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) "
-            "AppleWebKit/537.36 (KHTML, like Gecko) Antigravity/1.18.3"
+            "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Antigravity/1.18.3"
        ),
        "X-Goog-Api-Client": "google-cloud-sdk vscode_cloudshelleditor/0.1",
    }
@@ -316,9 +308,7 @@ def validate_credentials(access_token: str, project_id: str = _DEFAULT_PROJECT_I
        return False


-def refresh_access_token(
-    refresh_token: str, client_id: str, client_secret: str | None
-) -> dict | None:
+def refresh_access_token(refresh_token: str, client_id: str, client_secret: str | None) -> dict | None:
    """Refresh the access token using the refresh token."""
    data = {
        "grant_type": "refresh_token",
@@ -361,9 +351,7 @@ def cmd_account_add(args: argparse.Namespace) -> int:
        access_token = account.get("access")
        refresh_token_str = account.get("refresh", "")
        refresh_token = refresh_token_str.split("|")[0] if refresh_token_str else None
-        project_id = (
-            refresh_token_str.split("|")[1] if "|" in refresh_token_str else _DEFAULT_PROJECT_ID
-        )
+        project_id = refresh_token_str.split("|")[1] if "|" in refresh_token_str else _DEFAULT_PROJECT_ID
        email = account.get("email", "unknown")
        expires_ms = account.get("expires", 0)
        expires_at = expires_ms / 1000.0 if expires_ms else 0.0
@@ -390,9 +378,7 @@ def cmd_account_add(args: argparse.Namespace) -> int:
                    # Update the account
                    account["access"] = new_access
                    account["expires"] = int((time.time() + expires_in) * 1000)
-                    accounts_data["last_refresh"] = time.strftime(
-                        "%Y-%m-%dT%H:%M:%SZ", time.gmtime()
-                    )
+                    accounts_data["last_refresh"] = time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())
                    save_accounts(accounts_data)

                    # Validate the refreshed token
@@ -1,132 +0,0 @@
-"""
-Minimal Manual Agent Example
----------------------------
-This example demonstrates how to build and run an agent programmatically
-without using the Claude Code CLI or external LLM APIs.
-
-It uses custom NodeProtocol implementations to define logic in pure Python,
-making it perfect for understanding the core runtime loop:
-Setup -> Graph definition -> Execution -> Result
-
-Run with:
-    uv run python core/examples/manual_agent.py
-"""
-
-import asyncio
-
-from framework.graph import EdgeCondition, EdgeSpec, Goal, GraphSpec, NodeSpec
-from framework.graph.executor import GraphExecutor
-from framework.graph.node import NodeContext, NodeProtocol, NodeResult
-from framework.runtime.core import Runtime
-
-
-# 1. Define Node Logic (Custom NodeProtocol implementations)
-class GreeterNode(NodeProtocol):
-    """Generate a simple greeting."""
-
-    async def execute(self, ctx: NodeContext) -> NodeResult:
-        name = ctx.input_data.get("name", "World")
-        greeting = f"Hello, {name}!"
-        ctx.buffer.write("greeting", greeting)
-        return NodeResult(success=True, output={"greeting": greeting})
-
-
-class UppercaserNode(NodeProtocol):
-    """Convert text to uppercase."""
-
-    async def execute(self, ctx: NodeContext) -> NodeResult:
-        greeting = ctx.input_data.get("greeting") or ctx.buffer.read("greeting") or ""
-        result = greeting.upper()
-        ctx.buffer.write("final_greeting", result)
-        return NodeResult(success=True, output={"final_greeting": result})
-
-
-async def main():
-    print("Setting up Manual Agent...")
-
-    # 2. Define the Goal
-    # Every agent needs a goal with success criteria
-    goal = Goal(
-        id="greet-user",
-        name="Greet User",
-        description="Generate a friendly uppercase greeting",
-        success_criteria=[
-            {
-                "id": "greeting_generated",
-                "description": "Greeting produced",
-                "metric": "custom",
-                "target": "any",
-            }
-        ],
-    )
-
-    # 3. Define Nodes
-    # Nodes describe steps in the process
-    node1 = NodeSpec(
-        id="greeter",
-        name="Greeter",
-        description="Generates a simple greeting",
-        node_type="event_loop",
-        input_keys=["name"],
-        output_keys=["greeting"],
-    )
-
-    node2 = NodeSpec(
-        id="uppercaser",
-        name="Uppercaser",
-        description="Converts greeting to uppercase",
-        node_type="event_loop",
-        input_keys=["greeting"],
-        output_keys=["final_greeting"],
-    )
-
-    # 4. Define Edges
-    # Edges define the flow between nodes
-    edge1 = EdgeSpec(
-        id="greet-to-upper",
-        source="greeter",
-        target="uppercaser",
-        condition=EdgeCondition.ON_SUCCESS,
-    )
-
-    # 5. Create Graph
-    # The graph works like a blueprint connecting nodes and edges
-    graph = GraphSpec(
-        id="greeting-agent",
-        goal_id="greet-user",
-        entry_node="greeter",
-        terminal_nodes=["uppercaser"],
-        nodes=[node1, node2],
-        edges=[edge1],
-    )
-
-    # 6. Initialize Runtime & Executor
-    # Runtime handles state/memory; Executor runs the graph
-    from pathlib import Path
-
-    runtime = Runtime(storage_path=Path("./agent_logs"))
-    executor = GraphExecutor(runtime=runtime)
-
-    # 7. Register Node Implementations
-    # Connect node IDs in the graph to actual Python implementations
-    executor.register_node("greeter", GreeterNode())
-    executor.register_node("uppercaser", UppercaserNode())
-
-    # 8. Execute Agent
-    print("Executing agent with input: name='Alice'...")
-
-    result = await executor.execute(graph=graph, goal=goal, input_data={"name": "Alice"})
-
-    # 9. Verify Results
-    if result.success:
-        print("\nSuccess!")
-        print(f"Path taken: {' -> '.join(result.path)}")
-        print(f"Final output: {result.output.get('final_greeting')}")
-    else:
-        print(f"\nFailed: {result.error}")
-
-
-if __name__ == "__main__":
-    # Optional: Enable logging to see internal decision flow
-    # logging.basicConfig(level=logging.INFO)
-    asyncio.run(main())
@@ -1,119 +0,0 @@
-#!/usr/bin/env python3
-"""
-Example: Integrating MCP Servers with the Core Framework
-
-This example demonstrates how to:
-1. Register MCP servers programmatically
-2. Use MCP tools in agents
-3. Load MCP servers from configuration files
-"""
-
-import asyncio
-from pathlib import Path
-
-from framework.runner.runner import AgentRunner
-
-
-async def example_1_programmatic_registration():
-    """Example 1: Register MCP server programmatically"""
-    print("\n=== Example 1: Programmatic MCP Server Registration ===\n")
-
-    # Load an existing agent
-    runner = AgentRunner.load("exports/task-planner")
-
-    # Register tools MCP server via STDIO
-    num_tools = runner.register_mcp_server(
-        name="tools",
-        transport="stdio",
-        command="python",
-        args=["-m", "aden_tools.mcp_server", "--stdio"],
-        cwd="../tools",
-    )
-
-    print(f"Registered {num_tools} tools from tools MCP server")
-
-    # List all available tools
-    tools = runner._tool_registry.get_tools()
-    print(f"\nAvailable tools: {list(tools.keys())}")
-
-    # Run the agent with MCP tools available
-    result = await runner.run(
-        {"objective": "Search for 'Claude AI' and summarize the top 3 results"}
-    )
-
-    print(f"\nAgent result: {result}")
-
-    # Cleanup
-    runner.cleanup()
-
-
-async def example_2_http_transport():
-    """Example 2: Connect to MCP server via HTTP"""
-    print("\n=== Example 2: HTTP MCP Server Connection ===\n")
-
-    # First, start the tools MCP server in HTTP mode:
-    # cd tools && python mcp_server.py --port 4001
-
-    runner = AgentRunner.load("exports/task-planner")
-
-    # Register tools via HTTP
-    num_tools = runner.register_mcp_server(
-        name="tools-http",
-        transport="http",
-        url="http://localhost:4001",
-    )
-
-    print(f"Registered {num_tools} tools from HTTP MCP server")
-
-    # Cleanup
-    runner.cleanup()
-
-
-async def example_3_config_file():
-    """Example 3: Load MCP servers from configuration file"""
-    print("\n=== Example 3: Load from Configuration File ===\n")
-
-    # Create a test agent folder with mcp_servers.json
-    test_agent_path = Path("exports/task-planner")
-
-    # Copy example config (in practice, you'd place this in your agent folder)
-    import shutil
-
-    shutil.copy(Path(__file__).parent / "mcp_servers.json", test_agent_path / "mcp_servers.json")
-
-    # Load agent - MCP servers will be auto-discovered
-    runner = AgentRunner.load(test_agent_path)
-
-    # Tools are automatically available
-    tools = runner._tool_registry.get_tools()
-    print(f"Available tools: {list(tools.keys())}")
-
-    # Cleanup
-    runner.cleanup()
-
-    # Clean up the test config
-    (test_agent_path / "mcp_servers.json").unlink()
-
-
-async def main():
-    """Run all examples"""
-    print("=" * 60)
-    print("MCP Integration Examples")
-    print("=" * 60)
-
-    try:
-        # Run examples
-        await example_1_programmatic_registration()
-        # await example_2_http_transport()  # Requires HTTP server running
-        # await example_3_config_file()
-        # await example_4_custom_agent_with_mcp_tools()
-
-    except Exception as e:
-        print(f"\nError running example: {e}")
-        import traceback
-
-        traceback.print_exc()
-
-
-if __name__ == "__main__":
-    asyncio.run(main())
@@ -1,66 +1,20 @@
-"""
-Aden Hive Framework: A goal-driven agent runtime optimized for Builder observability.
+"""Hive Agent Framework.

-The runtime is designed around DECISIONS, not just actions. Every significant
-choice the agent makes is captured with:
- What it was trying to do (intent)
- What options it considered
- What it chose and why
- What happened as a result
- Whether that was good or bad (evaluated post-hoc)
-
-This gives the Builder LLM the information it needs to improve agent behavior.
-
-## Testing Framework
-
-The framework includes a Goal-Based Testing system (Goal → Agent → Eval):
- Generate tests from Goal success_criteria and constraints
- Mandatory user approval before tests are stored
- Parallel test execution with error categorization
- Debug tools with fix suggestions
-
-See `framework.testing` for details.
+Core classes:
+    ColonyRuntime -- orchestrates parallel worker clones in a colony
+    AgentLoop      -- the LLM + tool execution loop (one per worker)
+    AgentLoader    -- loads agent config from disk, builds pipeline
+    DecisionTracker -- records decisions for post-hoc analysis
 """

-from framework.llm import AnthropicProvider, LLMProvider
-from framework.runner import AgentRunner
-from framework.runtime.core import Runtime
-from framework.schemas.decision import Decision, DecisionEvaluation, Option, Outcome
-from framework.schemas.run import Problem, Run, RunSummary
-
-# Testing framework
-from framework.testing import (
-    ApprovalStatus,
-    DebugTool,
-    ErrorCategory,
-    Test,
-    TestResult,
-    TestStorage,
-    TestSuiteResult,
-)
+from framework.agent_loop import AgentLoop
+from framework.host import ColonyRuntime
+from framework.loader import AgentLoader
+from framework.tracker import DecisionTracker

 __all__ = [
-    # Schemas
-    "Decision",
-    "Option",
-    "Outcome",
-    "DecisionEvaluation",
-    "Run",
-    "RunSummary",
-    "Problem",
-    # Runtime
-    "Runtime",
-    # LLM
-    "LLMProvider",
-    "AnthropicProvider",
-    # Runner
-    "AgentRunner",
-    # Testing
-    "Test",
-    "TestResult",
-    "TestSuiteResult",
-    "TestStorage",
-    "ApprovalStatus",
-    "ErrorCategory",
-    "DebugTool",
+    "ColonyRuntime",
+    "AgentLoader",
+    "AgentLoop",
+    "DecisionTracker",
 ]
@@ -0,0 +1,34 @@
+"""Agent loop -- the core agent execution primitive."""
+
+from framework.agent_loop.conversation import (  # noqa: F401
+    ConversationStore,
+    Message,
+    NodeConversation,
+)
+from framework.agent_loop.types import (  # noqa: F401
+    AgentContext,
+    AgentProtocol,
+    AgentResult,
+    AgentSpec,
+)
+
+
+def __getattr__(name: str):
+    if name in ("AgentLoop", "JudgeProtocol", "JudgeVerdict", "LoopConfig", "OutputAccumulator"):
+        from framework.agent_loop.agent_loop import (
+            AgentLoop,
+            JudgeProtocol,
+            JudgeVerdict,
+            LoopConfig,
+            OutputAccumulator,
+        )
+
+        _exports = {
+            "AgentLoop": AgentLoop,
+            "JudgeProtocol": JudgeProtocol,
+            "JudgeVerdict": JudgeVerdict,
+            "LoopConfig": LoopConfig,
+            "OutputAccumulator": OutputAccumulator,
+        }
+        return _exports[name]
+    raise AttributeError(f"module {__name__!r} has no attribute {name!r}")
@@ -3,12 +3,14 @@
 from __future__ import annotations

 import json
+import logging
 import re
 from dataclasses import dataclass
 from pathlib import Path
 from typing import Any, Literal, Protocol, runtime_checkable

 LEGACY_RUN_ID = "__legacy_run__"
+logger = logging.getLogger(__name__)


 def is_legacy_run_id(run_id: str | None) -> bool:
@@ -46,6 +48,24 @@ class Message:
    is_skill_content: bool = False
    # Logical worker run identifier for shared-session persistence
    run_id: str | None = None
+    # True when this is a framework-injected continuation hint (continue-nudge
+    # on stream stall). Stored as a user message for API compatibility, but
+    # the UI should render it as a compact system notice, not user speech.
+    is_system_nudge: bool = False
+    # True when this message is a partial/truncated assistant turn reconstructed
+    # from a crashed or watchdog-cancelled stream. Signals that the original
+    # turn never finished — the model may or may not choose to redo it.
+    truncated: bool = False
+    # When non-None, identifies the parent session id this message was
+    # carried over from — used by fork_session_into_colony on the single
+    # compacted-summary message it writes when a colony is born from a
+    # queen DM. Presence of the field IS the "inherited" signal.
+    inherited_from: str | None = None
+    # True when this user message was synthesized from one or more
+    # fired triggers (timer/webhook), not typed by a human. The LLM still
+    # sees the message as a regular user turn; the UI uses this flag to
+    # render it as a trigger banner instead of a speech bubble.
+    is_trigger: bool = False

    def to_llm_dict(self) -> dict[str, Any]:
        """Convert to OpenAI-format message dict."""
@@ -59,9 +79,12 @@ class Message:
            return {"role": "user", "content": self.content}

        if self.role == "assistant":
-            d: dict[str, Any] = {"role": "assistant", "content": self.content}
+            d: dict[str, Any] = {"role": "assistant"}
            if self.tool_calls:
                d["tool_calls"] = self.tool_calls
+                d["content"] = self.content if self.content else None
+            else:
+                d["content"] = self.content or ""
            return d

        # role == "tool"
@@ -104,6 +127,14 @@ class Message:
            d["image_content"] = self.image_content
        if self.run_id is not None:
            d["run_id"] = self.run_id
+        if self.is_system_nudge:
+            d["is_system_nudge"] = self.is_system_nudge
+        if self.truncated:
+            d["truncated"] = self.truncated
+        if self.inherited_from is not None:
+            d["inherited_from"] = self.inherited_from
+        if self.is_trigger:
+            d["is_trigger"] = self.is_trigger
        return d

    @classmethod
@@ -121,6 +152,10 @@ class Message:
            is_client_input=data.get("is_client_input", False),
            image_content=data.get("image_content"),
            run_id=data.get("run_id"),
+            is_system_nudge=data.get("is_system_nudge", False),
+            truncated=data.get("truncated", False),
+            inherited_from=data.get("inherited_from"),
+            is_trigger=data.get("is_trigger", False),
        )


@@ -157,10 +192,17 @@ def update_run_cursor(
 def _extract_spillover_filename(content: str) -> str | None:
    """Extract spillover filename from a tool result annotation.

-    Matches patterns produced by EventLoopNode._truncate_tool_result():
-        - Large result:  "saved to 'web_search_1.txt'"
-        - Small result:  "[Saved to 'web_search_1.txt']"
+    Matches patterns produced by ``truncate_tool_result``:
+        - New large-result header: "Full result saved at: /abs/path/file.txt"
+        - Legacy bracketed trailer: "[Saved to 'file.txt']"  (pre-2026-04-15,
+          retained here so cold conversations still resolve)
    """
+    # New prose format — ``saved at: <absolute path>``, terminated by
+    # newline or end-of-string.
+    match = re.search(r"[Ss]aved at:\s*(\S+)", content)
+    if match:
+        return match.group(1)
+    # Legacy format.
    match = re.search(r"[Ss]aved to '([^']+)'", content)
    return match.group(1) if match else None

@@ -233,8 +275,8 @@ def extract_tool_call_history(messages: list[Message], max_entries: int = 30) ->
            return args.get("query", "")
        if name == "web_scrape":
            return args.get("url", "")
-        if name in ("load_data", "save_data"):
-            return args.get("filename", "")
+        if name == "read_file":
+            return args.get("path", "")
        return ""

    for msg in messages:
@@ -250,8 +292,8 @@ def extract_tool_call_history(messages: list[Message], max_entries: int = 30) ->
                summary = _summarize_input(name, args)
                tool_calls_detail.setdefault(name, []).append(summary)

-                if name == "save_data" and args.get("filename"):
-                    files_saved.append(args["filename"])
+                if name == "read_file" and args.get("path"):
+                    files_saved.append(args["path"])
                if name == "set_output" and args.get("key"):
                    outputs_set.append(args["key"])

@@ -305,6 +347,14 @@ class ConversationStore(Protocol):

    async def delete_parts_before(self, seq: int, run_id: str | None = None) -> None: ...

+    async def write_partial(self, seq: int, data: dict[str, Any]) -> None: ...
+
+    async def read_partial(self, seq: int) -> dict[str, Any] | None: ...
+
+    async def read_all_partials(self) -> list[dict[str, Any]]: ...
+
+    async def clear_partial(self, seq: int) -> None: ...
+
    async def close(self) -> None: ...

    async def destroy(self) -> None: ...
@@ -324,7 +374,7 @@ def _try_extract_key(content: str, key: str) -> str | None:
    3. Colon format: ``key: value``.
    4. Equals format: ``key = value``.
    """
-    from framework.graph.node import find_json_object
+    from framework.orchestrator.node import find_json_object

    # 1. Whole message is JSON
    try:
@@ -376,10 +426,36 @@ class NodeConversation:
        output_keys: list[str] | None = None,
        store: ConversationStore | None = None,
        run_id: str | None = None,
+        compaction_buffer_tokens: int | None = None,
+        compaction_buffer_ratio: float | None = None,
+        compaction_warning_buffer_tokens: int | None = None,
    ) -> None:
        self._system_prompt = system_prompt
+        # Optional split: when a caller updates the prompt with a
+        # ``dynamic_suffix`` argument, we remember the static prefix and
+        # suffix separately so the LLM wrapper can emit them as two
+        # Anthropic system content blocks with a cache breakpoint between
+        # them. ``_system_prompt`` stays as the concatenated form used for
+        # persistence and for the legacy single-block LLM path.
+        # On restore, these default to the concat/empty pair — the next
+        # AgentLoop iteration's dynamic-prompt refresh step repopulates.
+        self._system_prompt_static: str = system_prompt
+        self._system_prompt_dynamic_suffix: str = ""
        self._max_context_tokens = max_context_tokens
        self._compaction_threshold = compaction_threshold
+        # Buffer-based compaction trigger (Gap 7). When set, takes
+        # precedence over the multiplicative compaction_threshold so the
+        # loop reserves a fixed headroom for the next turn's input+output
+        # instead of trying to get exactly X% of the way to the hard
+        # limit. If left as None the legacy threshold-based rule is
+        # used, keeping old call sites behaving identically.
+        self._compaction_buffer_tokens = compaction_buffer_tokens
+        # Ratio component of the hybrid buffer. Combines additively with
+        # _compaction_buffer_tokens so callers can express "reserve N tokens
+        # plus M% of the window" — the absolute floor matters on tiny
+        # windows, the ratio matters on large ones.
+        self._compaction_buffer_ratio = compaction_buffer_ratio
+        self._compaction_warning_buffer_tokens = compaction_warning_buffer_tokens
        self._output_keys = output_keys
        self._store = store
        self._messages: list[Message] = []
@@ -393,15 +469,56 @@ class NodeConversation:

    @property
    def system_prompt(self) -> str:
+        """Full concatenated system prompt (static + dynamic suffix, if any).
+
+        This is the canonical form used for persistence and for the legacy
+        single-block LLM path. Split-prompt callers should read
+        ``system_prompt_static`` and ``system_prompt_dynamic_suffix`` instead.
+        """
        return self._system_prompt

-    def update_system_prompt(self, new_prompt: str) -> None:
+    @property
+    def system_prompt_static(self) -> str:
+        """Static prefix of the system prompt (cache-stable).
+
+        Equals ``system_prompt`` when no split is in use. When the AgentLoop
+        calls ``update_system_prompt(static, dynamic_suffix=...)``, this is
+        the piece sent as the cache-controlled first block.
+        """
+        return self._system_prompt_static
+
+    @property
+    def system_prompt_dynamic_suffix(self) -> str:
+        """Dynamic tail of the system prompt (not cached).
+
+        Empty unless the consumer splits its prompt. The LLM wrapper uses a
+        non-empty suffix to emit a two-block system content list with a
+        cache breakpoint between the static prefix and this tail.
+        """
+        return self._system_prompt_dynamic_suffix
+
+    def update_system_prompt(self, new_prompt: str, dynamic_suffix: str | None = None) -> None:
        """Update the system prompt.

        Used in continuous conversation mode at phase transitions to swap
        Layer 3 (focus) while preserving the conversation history.
+
+        When ``dynamic_suffix`` is provided, ``new_prompt`` is interpreted as
+        the STATIC prefix and ``dynamic_suffix`` as the per-turn tail; they
+        travel to the LLM as two separate cache-controlled blocks but are
+        persisted as a single concatenated string for backward-compat
+        restore. ``new_prompt`` alone (suffix left None) keeps the legacy
+        single-string behavior.
        """
-        self._system_prompt = new_prompt
+        if dynamic_suffix is None:
+            # Legacy single-string path — static == full, no suffix split.
+            self._system_prompt = new_prompt
+            self._system_prompt_static = new_prompt
+            self._system_prompt_dynamic_suffix = ""
+        else:
+            self._system_prompt_static = new_prompt
+            self._system_prompt_dynamic_suffix = dynamic_suffix
+            self._system_prompt = f"{new_prompt}\n\n{dynamic_suffix}" if dynamic_suffix else new_prompt
        self._meta_persisted = False  # re-persist with new prompt

    def set_current_phase(self, phase_id: str) -> None:
@@ -440,6 +557,8 @@ class NodeConversation:
        is_transition_marker: bool = False,
        is_client_input: bool = False,
        image_content: list[dict[str, Any]] | None = None,
+        is_system_nudge: bool = False,
+        is_trigger: bool = False,
    ) -> Message:
        msg = Message(
            seq=self._next_seq,
@@ -450,6 +569,8 @@ class NodeConversation:
            is_transition_marker=is_transition_marker,
            is_client_input=is_client_input,
            image_content=image_content,
+            is_system_nudge=is_system_nudge,
+            is_trigger=is_trigger,
        )
        self._messages.append(msg)
        self._next_seq += 1
@@ -463,6 +584,8 @@ class NodeConversation:
        self,
        content: str,
        tool_calls: list[dict[str, Any]] | None = None,
+        *,
+        truncated: bool = False,
    ) -> Message:
        msg = Message(
            seq=self._next_seq,
@@ -471,6 +594,7 @@ class NodeConversation:
            tool_calls=tool_calls,
            phase_id=self._current_phase,
            run_id=self._run_id,
+            truncated=truncated,
        )
        self._messages.append(msg)
        self._next_seq += 1
@@ -486,6 +610,27 @@ class NodeConversation:
        image_content: list[dict[str, Any]] | None = None,
        is_skill_content: bool = False,
    ) -> Message:
+        # Dedup guard: reject a second tool_result for the same tool_use_id.
+        # Anthropic's API only accepts one result per tool_call, and a duplicate
+        # causes a hard 400 two turns later ("messages with role 'tool' must
+        # be a response to a preceding message with 'tool_calls'"). Duplicates
+        # can arise when a tool_call_timeout fires and records a placeholder
+        # error, then the real executor thread eventually delivers the actual
+        # result (the thread kept running inside run_in_executor — see
+        # tool_result_handler.execute_tool).  We keep the FIRST result to
+        # preserve whatever state the agent already reasoned about.
+        for existing in reversed(self._messages):
+            if existing.role == "tool" and existing.tool_use_id == tool_use_id:
+                import logging as _logging
+
+                _logging.getLogger(__name__).warning(
+                    "add_tool_result: dropping duplicate result for tool_use_id=%s "
+                    "(first result preserved, %d chars; new result ignored, %d chars)",
+                    tool_use_id,
+                    len(existing.content),
+                    len(content),
+                )
+                return existing
        msg = Message(
            seq=self._next_seq,
            role="tool",
@@ -505,6 +650,59 @@ class NodeConversation:

    # --- Query -------------------------------------------------------------

+    def find_completed_tool_call(
+        self,
+        name: str,
+        tool_input: dict[str, Any],
+        within_last_turns: int = 3,
+    ) -> Message | None:
+        """Return the most recent assistant message that issued a tool call
+        with the same (name + canonical-json args) AND received a non-error
+        tool result, within the last ``within_last_turns`` assistant turns.
+
+        Used by the replay detector to flag when the model is about to redo
+        a successful call — we prepend a steer onto the upcoming result but
+        still execute, so tools like browser_screenshot that are legitimately
+        repeated are not silently skipped.
+        """
+        try:
+            target_canonical = json.dumps(tool_input, sort_keys=True, default=str)
+        except (TypeError, ValueError):
+            target_canonical = str(tool_input)
+
+        # Walk backwards over recent assistant messages
+        assistant_turns_seen = 0
+        for idx in range(len(self._messages) - 1, -1, -1):
+            m = self._messages[idx]
+            if m.role != "assistant":
+                continue
+            assistant_turns_seen += 1
+            if assistant_turns_seen > within_last_turns:
+                break
+            if not m.tool_calls:
+                continue
+            for tc in m.tool_calls:
+                func = tc.get("function", {}) if isinstance(tc, dict) else {}
+                tc_name = func.get("name")
+                if tc_name != name:
+                    continue
+                args_str = func.get("arguments", "")
+                try:
+                    parsed = json.loads(args_str) if isinstance(args_str, str) else args_str
+                    canonical = json.dumps(parsed, sort_keys=True, default=str)
+                except (TypeError, ValueError):
+                    canonical = str(args_str)
+                if canonical != target_canonical:
+                    continue
+                # Found a match — now verify its result was not an error.
+                tc_id = tc.get("id")
+                for later in self._messages[idx + 1 :]:
+                    if later.role == "tool" and later.tool_use_id == tc_id:
+                        if not later.is_error:
+                            return m
+                        break
+        return None
+
    def to_llm_messages(self) -> list[dict[str, Any]]:
        """Return messages as OpenAI-format dicts (system prompt excluded).

@@ -513,7 +711,48 @@ class NodeConversation:
        can happen when a loop is cancelled mid-tool-execution.
        """
        msgs = [m.to_llm_dict() for m in self._messages]
-        return self._repair_orphaned_tool_calls(msgs)
+        msgs = self._repair_orphaned_tool_calls(msgs)
+        msgs = self._sanitize_for_api(msgs)
+        return msgs
+
+    @staticmethod
+    def _sanitize_for_api(msgs: list[dict[str, Any]]) -> list[dict[str, Any]]:
+        """Final pass: ensure message sequence is valid for strict APIs.
+
+        Rules:
+        1. No two consecutive messages with the same role (merge or drop)
+        2. Tool messages must have a tool_call_id
+        3. Assistant messages with tool_calls must have content=null, not ""
+        4. First message must not be 'tool' or 'assistant' (without prior context)
+        """
+        cleaned: list[dict[str, Any]] = []
+        for m in msgs:
+            role = m.get("role")
+
+            # Fix assistant content when tool_calls present
+            if role == "assistant" and m.get("tool_calls"):
+                if m.get("content") == "":
+                    m["content"] = None
+
+            # Drop tool messages without tool_call_id
+            if role == "tool" and not m.get("tool_call_id"):
+                continue
+
+            # Drop consecutive duplicate roles (merge user messages)
+            if cleaned and cleaned[-1].get("role") == role == "user":
+                prev_content = cleaned[-1].get("content", "")
+                curr_content = m.get("content", "")
+                if isinstance(prev_content, str) and isinstance(curr_content, str):
+                    cleaned[-1]["content"] = f"{prev_content}\n{curr_content}"
+                    continue
+
+            cleaned.append(m)
+
+        # Drop leading assistant/tool messages (no prior context)
+        while cleaned and cleaned[0].get("role") in ("assistant", "tool"):
+            cleaned.pop(0)
+
+        return cleaned

    @staticmethod
    def _repair_orphaned_tool_calls(
@@ -521,11 +760,18 @@ class NodeConversation:
    ) -> list[dict[str, Any]]:
        """Ensure tool_call / tool_result pairs are consistent.

-        1. **Orphaned tool results** (tool_result with no preceding tool_use)
-           are dropped.  This happens when compaction removes an assistant
-           message but leaves its tool-result messages behind.
-        2. **Orphaned tool calls** (tool_use with no following tool_result)
-           get a synthetic error result appended.  This happens when a loop
+        1. **Orphaned tool results** (tool_result with no matching tool_use
+           anywhere) are dropped.  Happens after compaction removes the
+           parent assistant message.
+        2. **Positionally orphaned tool results** (tool_result separated
+           from its parent by a non-tool message, e.g. a user injection)
+           are dropped.  The Anthropic API requires tool messages to
+           follow immediately after the assistant message that issued
+           the matching tool_call.
+        3. **Duplicate tool results** (same tool_call_id appearing more
+           than once) are dropped; only the first is kept.
+        4. **Orphaned tool calls** (tool_use with no following tool_result)
+           get a synthetic error result appended.  Happens when the loop
           is cancelled mid-tool-execution.
        """
        # Pass 1: collect all tool_call IDs from assistant messages so we
@@ -538,41 +784,75 @@ class NodeConversation:
                    if tc_id:
                        all_tool_call_ids.add(tc_id)

-        # Pass 2: build repaired list — drop orphaned tool results, patch
-        # missing tool results.
+        # Pass 2: build repaired list — drop orphaned tool results, drop
+        # positional orphans and duplicates, patch missing tool results.
+        #
+        # ``open_tool_calls`` holds the tool_call IDs we're still expecting
+        # results for: it's populated when we emit an assistant-with-tool_calls
+        # and drained as matching tool messages follow. Any tool message
+        # whose id is not currently open is positionally invalid and gets
+        # dropped — that closes the gap that caused the tool-after-user
+        # 400 errors.
        repaired: list[dict[str, Any]] = []
-        for i, m in enumerate(msgs):
-            # Drop tool-result messages whose tool_call_id has no matching
-            # tool_use in any assistant message (orphaned by compaction).
-            if m.get("role") == "tool":
-                tid = m.get("tool_call_id")
-                if tid and tid not in all_tool_call_ids:
-                    continue  # skip orphaned result
+        open_tool_calls: set[str] = set()
+        seen_tool_ids: set[str] = set()
+        for m in msgs:
+            role = m.get("role")

-            repaired.append(m)
-            tool_calls = m.get("tool_calls")
-            if m.get("role") != "assistant" or not tool_calls:
+            if role == "tool":
+                tid = m.get("tool_call_id")
+                # Drop tool results with no matching tool_use anywhere.
+                if not tid or tid not in all_tool_call_ids:
+                    continue
+                # Drop duplicates (same id appearing twice) — keep first.
+                if tid in seen_tool_ids:
+                    continue
+                # Drop positional orphans — tool messages whose parent
+                # assistant isn't the still-open assistant block.
+                if tid not in open_tool_calls:
+                    continue
+                open_tool_calls.discard(tid)
+                seen_tool_ids.add(tid)
+                repaired.append(m)
                continue
-            # Collect IDs of tool results that follow this assistant message
-            answered: set[str] = set()
-            for j in range(i + 1, len(msgs)):
-                if msgs[j].get("role") == "tool":
-                    tid = msgs[j].get("tool_call_id")
-                    if tid:
-                        answered.add(tid)
-                else:
-                    break  # stop at first non-tool message
-            # Patch any missing results
-            for tc in tool_calls:
-                tc_id = tc.get("id")
-                if tc_id and tc_id not in answered:
+
+            # Any non-tool message closes the current assistant tool block.
+            # If the previous assistant left tool_calls unanswered, patch
+            # synthetic error results before emitting this message so the
+            # API sees a complete pairing.
+            if open_tool_calls:
+                for stale_id in list(open_tool_calls):
                    repaired.append(
                        {
                            "role": "tool",
-                            "tool_call_id": tc_id,
+                            "tool_call_id": stale_id,
                            "content": "ERROR: Tool execution was interrupted.",
                        }
                    )
+                    seen_tool_ids.add(stale_id)
+                open_tool_calls.clear()
+
+            repaired.append(m)
+
+            if role == "assistant":
+                for tc in m.get("tool_calls") or []:
+                    tc_id = tc.get("id")
+                    if tc_id and tc_id not in seen_tool_ids:
+                        open_tool_calls.add(tc_id)
+
+        # Tail: if the conversation ends with an assistant that issued
+        # tool_calls and no results followed, patch them so the next
+        # turn's first message can be a valid assistant/user response.
+        if open_tool_calls:
+            for stale_id in list(open_tool_calls):
+                repaired.append(
+                    {
+                        "role": "tool",
+                        "tool_call_id": stale_id,
+                        "content": "ERROR: Tool execution was interrupted.",
+                    }
+                )
+
        return repaired

    def estimate_tokens(self) -> int:
@@ -621,8 +901,48 @@ class NodeConversation:
        return self.estimate_tokens() / self._max_context_tokens

    def needs_compaction(self) -> bool:
+        """True when the conversation should be compacted before the
+        next LLM call.
+
+        Hybrid buffer rule: the headroom reserved before compaction fires
+        is the SUM of an absolute fixed component and a ratio of the hard
+        context limit:
+
+            effective_buffer = compaction_buffer_tokens
+                             + compaction_buffer_ratio * max_context_tokens
+
+        The fixed component gives a floor on tiny windows; the ratio
+        keeps the trigger meaningful on large windows where any constant
+        buffer becomes a rounding error (an 8k buffer is 75% on a 32k
+        window but 96% on a 200k window). Compaction fires when the
+        current estimate would consume more than (limit - effective_buffer).
+
+        When neither component is configured, falls back to the legacy
+        multiplicative threshold so old callers keep behaving identically.
+        """
+        if self._max_context_tokens <= 0:
+            return False
+        fixed = self._compaction_buffer_tokens
+        ratio = self._compaction_buffer_ratio
+        if fixed is not None or ratio is not None:
+            effective_buffer = (fixed or 0) + (ratio or 0.0) * self._max_context_tokens
+            budget = self._max_context_tokens - effective_buffer
+            return self.estimate_tokens() >= max(0.0, budget)
        return self.estimate_tokens() >= self._max_context_tokens * self._compaction_threshold

+    def compaction_warning(self) -> bool:
+        """True when the conversation has crossed the warning threshold
+        but not yet the hard compaction trigger.
+
+        Used by telemetry / UI to show a "context getting tight" hint
+        before a compaction pass actually runs. Returns False when no
+        warning buffer is configured (legacy behaviour).
+        """
+        if self._max_context_tokens <= 0 or self._compaction_warning_buffer_tokens is None:
+            return False
+        warn_at = self._max_context_tokens - self._compaction_warning_buffer_tokens
+        return self.estimate_tokens() >= max(0, warn_at)
+
    # --- Output-key extraction ---------------------------------------------

    def _extract_protected_values(self, messages: list[Message]) -> dict[str, str]:
@@ -699,7 +1019,7 @@ class NodeConversation:
                continue  # never prune errors
            if msg.is_skill_content:
                continue  # never prune activated skill instructions (AS-10)
-            if msg.content.startswith("[Pruned tool result"):
+            if msg.content.startswith(("Pruned tool result", "[Pruned tool result")):
                continue  # already pruned
            # Tiny results (set_output acks, confirmations) — pruning
            # saves negligible space but makes the LLM think the call
@@ -731,12 +1051,12 @@ class NodeConversation:

            if spillover:
                placeholder = (
-                    f"[Pruned tool result: {orig_len} chars. "
-                    f"Full data in '{spillover}'. "
-                    f"Use load_data('{spillover}') to retrieve.]"
+                    f"Pruned tool result ({orig_len:,} chars) cleared from context. "
+                    f"Full data saved at: {spillover}\n"
+                    f"Read the complete data with read_file(path='{spillover}')."
                )
            else:
-                placeholder = f"[Pruned tool result: {orig_len} chars cleared from context.]"
+                placeholder = f"Pruned tool result ({orig_len:,} chars) cleared from context."

            self._messages[i] = Message(
                seq=msg.seq,
@@ -758,6 +1078,78 @@ class NodeConversation:
        self._last_api_input_tokens = None
        return count

+    async def evict_old_images(self, keep_latest: int = 2) -> int:
+        """Strip ``image_content`` from older messages, keeping the most recent.
+
+        Screenshots from ``browser_screenshot`` are inlined into the
+        message's ``image_content`` as base64 data URLs. Each screenshot
+        costs ~250k tokens when the provider counts the base64 as
+        text — four screenshots push a conversation over gemini's 1M
+        context limit and trigger out-of-context garbage output (see
+        ``session_20260415_104727_5c4ed7ff`` for the terminal case
+        where the model emitted ``协日`` as its final text then stopped).
+
+        This method walks backward through messages and keeps
+        ``image_content`` intact on the most recent ``keep_latest``
+        messages that have images. Older messages get their
+        ``image_content`` nulled out — the text content (metadata
+        like url, dimensions, scale hints) stays, but the raw bytes
+        are dropped. Storage is updated too so cold-restore sees the
+        same evicted state.
+
+        Run this right after every tool result is recorded so image
+        context stays bounded even within a single iteration (the
+        compaction pipeline only fires at iteration boundaries, too
+        late for a single turn that takes 4 screenshots).
+
+        Returns the number of messages whose image_content was evicted.
+        """
+        if not self._messages or keep_latest < 0:
+            return 0
+
+        # Find messages carrying images, walking newest → oldest.
+        image_indices: list[int] = []
+        for i in range(len(self._messages) - 1, -1, -1):
+            if self._messages[i].image_content:
+                image_indices.append(i)
+
+        # Nothing to evict if we have ≤ keep_latest images total.
+        if len(image_indices) <= keep_latest:
+            return 0
+
+        # Evict everything past the first keep_latest (newest) entries.
+        to_evict = image_indices[keep_latest:]
+        evicted = 0
+        for idx in to_evict:
+            msg = self._messages[idx]
+            self._messages[idx] = Message(
+                seq=msg.seq,
+                role=msg.role,
+                content=msg.content,
+                tool_use_id=msg.tool_use_id,
+                tool_calls=msg.tool_calls,
+                is_error=msg.is_error,
+                phase_id=msg.phase_id,
+                is_transition_marker=msg.is_transition_marker,
+                is_client_input=msg.is_client_input,
+                image_content=None,  # ← dropped
+                is_skill_content=msg.is_skill_content,
+                run_id=msg.run_id,
+            )
+            evicted += 1
+            if self._store:
+                await self._store.write_part(msg.seq, self._messages[idx].to_storage_dict())
+
+        if evicted:
+            # Reset token estimate — image blocks no longer contribute.
+            self._last_api_input_tokens = None
+            logger.info(
+                "evict_old_images: dropped image_content from %d message(s), kept %d most recent",
+                evicted,
+                keep_latest,
+            )
+        return evicted
+
    async def compact(
        self,
        summary: str,
@@ -910,9 +1302,7 @@ class NodeConversation:
            for msg in old_messages:
                if msg.role != "assistant" or not msg.tool_calls:
                    continue
-                has_protected = any(
-                    tc.get("function", {}).get("name") == "set_output" for tc in msg.tool_calls
-                )
+                has_protected = any(tc.get("function", {}).get("name") == "set_output" for tc in msg.tool_calls)
                tc_ids = {tc.get("id", "") for tc in msg.tool_calls}
                if has_protected:
                    protected_tc_ids |= tc_ids
@@ -1018,16 +1408,18 @@ class NodeConversation:
            # Nothing to save — skip file creation
            conv_filename = ""

-        # Build reference message
+        # Build reference message. Prose format (no brackets) — see the
+        # poison-pattern note on truncate_tool_result. Frontier models
+        # autocomplete `[...']` trailers into their own text turns.
        ref_parts: list[str] = []
        if conv_filename:
            full_path = str((spill_path / conv_filename).resolve())
            ref_parts.append(
-                f"[Previous conversation saved to '{full_path}'. "
-                f"Use load_data('{conv_filename}') to review if needed.]"
+                f"Previous conversation saved at: {full_path}\n"
+                f"Read the full transcript with read_file('{conv_filename}')."
            )
        elif not collapsed_msgs:
-            ref_parts.append("[Previous freeform messages compacted.]")
+            ref_parts.append("(Previous freeform messages compacted.)")

        # Aggressive: add collapsed tool-call history to the reference
        if collapsed_msgs:
@@ -1106,11 +1498,7 @@ class NodeConversation:

    def export_summary(self) -> str:
        """Structured summary with [STATS], [CONFIG], [RECENT_MESSAGES] sections."""
-        prompt_preview = (
-            self._system_prompt[:80] + "..."
-            if len(self._system_prompt) > 80
-            else self._system_prompt
-        )
+        prompt_preview = self._system_prompt[:80] + "..." if len(self._system_prompt) > 80 else self._system_prompt

        lines = [
            "[STATS]",
@@ -1143,6 +1531,45 @@ class NodeConversation:
            await self._persist_meta()
        await self._store.write_part(message.seq, message.to_storage_dict())
        await self._write_next_seq()
+        # Any partial checkpoint for this seq is now superseded by the real
+        # part — clear it so a future restore doesn't resurrect stale text.
+        try:
+            await self._store.clear_partial(message.seq)
+        except AttributeError:
+            # Older stores may not implement partials; ignore.
+            pass
+
+    async def checkpoint_partial_assistant(
+        self,
+        accumulated_text: str,
+        tool_calls: list[dict[str, Any]] | None = None,
+    ) -> None:
+        """Write an in-flight assistant turn's state to disk under the next seq.
+
+        Called from the stream event loop. Safe to call repeatedly — each call
+        overwrites the prior checkpoint. Persisted via ``write_partial`` so it
+        does NOT appear in ``read_parts()`` and cannot be double-loaded. Cleared
+        automatically when ``add_assistant_message`` for this seq lands.
+        """
+        if self._store is None:
+            return
+        if not self._meta_persisted:
+            await self._persist_meta()
+        payload: dict[str, Any] = {
+            "seq": self._next_seq,
+            "role": "assistant",
+            "content": accumulated_text,
+            "phase_id": self._current_phase,
+            "run_id": self._run_id,
+            "truncated": True,
+        }
+        if tool_calls:
+            payload["tool_calls"] = tool_calls
+        try:
+            await self._store.write_partial(self._next_seq, payload)
+        except AttributeError:
+            # Older stores may not implement partials; ignore.
+            pass

    async def _persist_meta(self) -> None:
        """Lazily write conversation metadata to the store (called once).
@@ -1156,6 +1583,9 @@ class NodeConversation:
            "system_prompt": self._system_prompt,
            "max_context_tokens": self._max_context_tokens,
            "compaction_threshold": self._compaction_threshold,
+            "compaction_buffer_tokens": self._compaction_buffer_tokens,
+            "compaction_buffer_ratio": self._compaction_buffer_ratio,
+            "compaction_warning_buffer_tokens": (self._compaction_warning_buffer_tokens),
            "output_keys": self._output_keys,
        }
        await self._store.write_meta(run_meta)
@@ -1203,12 +1633,28 @@ class NodeConversation:
            output_keys=meta.get("output_keys"),
            store=store,
            run_id=run_id,
+            compaction_buffer_tokens=meta.get("compaction_buffer_tokens"),
+            compaction_buffer_ratio=meta.get("compaction_buffer_ratio"),
+            compaction_warning_buffer_tokens=meta.get("compaction_warning_buffer_tokens"),
        )
        conv._meta_persisted = True

        parts = await store.read_parts()
        if phase_id:
-            parts = [p for p in parts if p.get("phase_id") == phase_id]
+            filtered_parts = [p for p in parts if p.get("phase_id") == phase_id]
+            if filtered_parts:
+                parts = filtered_parts
+            elif parts and all(p.get("phase_id") is None for p in parts):
+                # Backward compatibility: older isolated stores (including queen
+                # sessions) persisted parts without phase_id. In that case, the
+                # phase filter would incorrectly hide the entire conversation.
+                logger.info(
+                    "Restoring legacy unphased conversation without applying phase filter (phase_id=%s, parts=%d)",
+                    phase_id,
+                    len(parts),
+                )
+            else:
+                parts = filtered_parts
        # Filter by run_id so intentional restarts (new run_id) start fresh
        # while crash recovery (same run_id) loads prior parts.
        if run_id and not is_legacy_run_id(run_id):
@@ -1222,4 +1668,45 @@ class NodeConversation:
        elif conv._messages:
            conv._next_seq = conv._messages[-1].seq + 1

+        # Surface any leftover partial checkpoints as truncated messages so
+        # the next turn sees what the interrupted stream was in the middle
+        # of producing. Only partials whose seq is >= next_seq are meaningful;
+        # anything lower was already superseded by a real part.
+        try:
+            partials = await store.read_all_partials()
+        except AttributeError:
+            partials = []
+        for p in partials:
+            pseq = p.get("seq", -1)
+            if pseq < conv._next_seq:
+                # Stale — clean it up.
+                try:
+                    await store.clear_partial(pseq)
+                except AttributeError:
+                    pass
+                continue
+            # Only resurrect partials relevant to this run / phase.
+            if run_id and not is_legacy_run_id(run_id) and p.get("run_id") != run_id:
+                continue
+            if phase_id and p.get("phase_id") is not None and p.get("phase_id") != phase_id:
+                continue
+            # Reconstruct as a truncated assistant message.
+            msg = Message(
+                seq=pseq,
+                role="assistant",
+                content=p.get("content", "") or "",
+                tool_calls=p.get("tool_calls"),
+                phase_id=p.get("phase_id"),
+                run_id=p.get("run_id"),
+                truncated=True,
+            )
+            conv._messages.append(msg)
+            conv._next_seq = max(conv._next_seq, pseq + 1)
+            logger.info(
+                "restore: resurrected truncated partial seq=%d (text=%d chars, tool_calls=%d)",
+                pseq,
+                len(msg.content),
+                len(msg.tool_calls or []),
+            )
+
        return conv
@@ -0,0 +1,7 @@
+"""Agent loop internals -- compaction, judge, tools, subagent execution.
+
+Re-exports from legacy locations for the new import path.
+"""
+
+from framework.agent_loop.internals.compaction import *  # noqa: F401, F403
+from framework.agent_loop.internals.synthetic_tools import *  # noqa: F401, F403
@@ -19,11 +19,11 @@ from datetime import UTC, datetime
 from pathlib import Path
 from typing import Any

-from framework.graph.conversation import Message, NodeConversation
-from framework.graph.event_loop.event_publishing import publish_context_usage
-from framework.graph.event_loop.types import LoopConfig, OutputAccumulator
-from framework.graph.node import NodeContext
-from framework.runtime.event_bus import EventBus
+from framework.agent_loop.conversation import Message, NodeConversation
+from framework.agent_loop.internals.event_publishing import publish_context_usage
+from framework.agent_loop.internals.types import LoopConfig, OutputAccumulator
+from framework.host.event_bus import EventBus
+from framework.orchestrator.node import NodeContext

 logger = logging.getLogger(__name__)

@@ -80,7 +80,7 @@ def microcompact(
        msg = messages[i]
        if msg.role != "tool" or msg.is_error or msg.is_skill_content:
            continue
-        if msg.content.startswith(("[Pruned tool result", "[Old tool result")):
+        if msg.content.startswith(("Pruned tool result", "[Pruned tool result", "[Old tool result")):
            continue
        if len(msg.content) < 100:
            continue
@@ -102,12 +102,12 @@ def microcompact(
        orig_len = len(msg.content)
        if spillover:
            placeholder = (
-                f"[Old tool result cleared: {orig_len} chars. "
-                f"Full data in '{spillover}'. "
-                f"Use load_data('{spillover}') to retrieve.]"
+                f"Old tool result ({orig_len:,} chars) cleared from context. "
+                f"Full data saved at: {spillover}\n"
+                f"Read the complete data with read_file(path='{spillover}')."
            )
        else:
-            placeholder = f"[Old tool result cleared: {orig_len} chars.]"
+            placeholder = f"Old tool result ({orig_len:,} chars) cleared from context."

        # Mutate in-place (microcompact is synchronous, no store writes)
        conversation._messages[i] = Message(
@@ -142,7 +142,14 @@ def _find_tool_name_for_result(messages: list[Message], tool_msg: Message) -> st


 def _extract_spillover_filename_inline(content: str) -> str | None:
-    """Quick inline check for spillover filename in tool result content."""
+    """Quick inline check for spillover filename in tool result content.
+
+    Matches both the new prose format ("saved at: /path") and the
+    legacy bracketed trailer ("saved to '/path'").
+    """
+    match = re.search(r"saved at:\s*(\S+)", content, re.IGNORECASE)
+    if match:
+        return match.group(1)
    match = re.search(r"saved to '([^']+)'", content, re.IGNORECASE)
    return match.group(1) if match else None

@@ -168,13 +175,17 @@ async def compact(
    """
    conv_id = id(conversation)

-    # Circuit breaker: stop auto-compacting after repeated failures
-    if _failure_counts.get(conv_id, 0) >= MAX_CONSECUTIVE_FAILURES:
+    # Circuit breaker: stop LLM-based compaction after repeated failures,
+    # but still fall through to the emergency deterministic summary so
+    # the conversation doesn't silently grow past the context window.
+    # Without this, a persistent LLM outage during compaction would
+    # leave the agent stuck sending oversized prompts until the API 400s.
+    _llm_compaction_skipped = _failure_counts.get(conv_id, 0) >= MAX_CONSECUTIVE_FAILURES
+    if _llm_compaction_skipped:
        logger.warning(
-            "Circuit breaker: skipping compaction after %d consecutive failures",
+            "Circuit breaker: LLM compaction disabled after %d failures — skipping straight to emergency summary",
            _failure_counts[conv_id],
        )
-        return

    # Recompaction detection
    now = time.monotonic()
@@ -256,7 +267,7 @@ async def compact(
        return

    # --- Step 3: LLM summary compaction ---
-    if ctx.llm is not None:
+    if ctx.llm is not None and not _llm_compaction_skipped:
        logger.info(
            "LLM summary compaction triggered (%.0f%% usage)",
            conversation.usage_ratio() * 100,
@@ -360,6 +371,7 @@ async def llm_compact(
    char_limit: int = LLM_COMPACT_CHAR_LIMIT,
    max_depth: int = LLM_COMPACT_MAX_DEPTH,
    max_context_tokens: int = 128_000,
+    preserve_user_messages: bool = False,
 ) -> str:
    """Summarise *messages* with LLM, splitting recursively if too large.

@@ -367,9 +379,14 @@ async def llm_compact(
    rejects the call with a context-length error, the messages are split
    in half and each half is summarised independently.  Tool history is
    appended once at the top-level call (``_depth == 0``).
+
+    When ``preserve_user_messages`` is True, the prompt and system message
+    are amplified to instruct the LLM to keep every user message verbatim
+    and in full — used by the manual /compact-and-fork endpoint where the
+    user wants their voice carried into the new session intact.
    """
-    from framework.graph.conversation import extract_tool_call_history
-    from framework.graph.event_loop.tool_result_handler import is_context_too_large_error
+    from framework.agent_loop.conversation import extract_tool_call_history
+    from framework.agent_loop.internals.tool_result_handler import is_context_too_large_error

    if _depth > max_depth:
        raise RuntimeError(f"LLM compaction recursion limit ({max_depth})")
@@ -390,6 +407,7 @@ async def llm_compact(
            char_limit=char_limit,
            max_depth=max_depth,
            max_context_tokens=max_context_tokens,
+            preserve_user_messages=preserve_user_messages,
        )
    else:
        prompt = build_llm_compaction_prompt(
@@ -397,17 +415,30 @@ async def llm_compact(
            accumulator,
            formatted,
            max_context_tokens=max_context_tokens,
+            preserve_user_messages=preserve_user_messages,
        )
+        if preserve_user_messages:
+            system_msg = (
+                "You are a conversation compactor for an AI agent. "
+                "Write a detailed summary that allows the agent to "
+                "continue its work. CRITICAL: reproduce every user "
+                "message verbatim and in full inside the 'User Messages' "
+                "section — do not paraphrase, truncate, or merge them. "
+                "Assistant turns and tool results may be summarised, but "
+                "user input is sacred."
+            )
+        else:
+            system_msg = (
+                "You are a conversation compactor for an AI agent. "
+                "Write a detailed summary that allows the agent to "
+                "continue its work. Preserve user-stated rules, "
+                "constraints, and account/identity preferences verbatim."
+            )
        summary_budget = max(1024, max_context_tokens // 2)
        try:
            response = await ctx.llm.acomplete(
                messages=[{"role": "user", "content": prompt}],
-                system=(
-                    "You are a conversation compactor for an AI agent. "
-                    "Write a detailed summary that allows the agent to "
-                    "continue its work. Preserve user-stated rules, "
-                    "constraints, and account/identity preferences verbatim."
-                ),
+                system=system_msg,
                max_tokens=summary_budget,
            )
            summary = response.content
@@ -426,6 +457,7 @@ async def llm_compact(
                    char_limit=char_limit,
                    max_depth=max_depth,
                    max_context_tokens=max_context_tokens,
+                    preserve_user_messages=preserve_user_messages,
                )
            else:
                raise
@@ -448,6 +480,7 @@ async def _llm_compact_split(
    char_limit: int = LLM_COMPACT_CHAR_LIMIT,
    max_depth: int = LLM_COMPACT_MAX_DEPTH,
    max_context_tokens: int = 128_000,
+    preserve_user_messages: bool = False,
 ) -> str:
    """Split messages in half and summarise each half independently."""
    mid = max(1, len(messages) // 2)
@@ -459,6 +492,7 @@ async def _llm_compact_split(
        char_limit=char_limit,
        max_depth=max_depth,
        max_context_tokens=max_context_tokens,
+        preserve_user_messages=preserve_user_messages,
    )
    s2 = await llm_compact(
        ctx,
@@ -468,6 +502,7 @@ async def _llm_compact_split(
        char_limit=char_limit,
        max_depth=max_depth,
        max_context_tokens=max_context_tokens,
+        preserve_user_messages=preserve_user_messages,
    )
    return s1 + "\n\n" + s2

@@ -499,6 +534,7 @@ def build_llm_compaction_prompt(
    formatted_messages: str,
    *,
    max_context_tokens: int = 128_000,
+    preserve_user_messages: bool = False,
 ) -> str:
    """Build prompt for LLM compaction targeting 50% of token budget.

@@ -506,7 +542,7 @@ def build_llm_compaction_prompt(
    service.  Each section focuses on a different aspect of the conversation
    so the summariser produces consistently useful, well-organised output.
    """
-    spec = ctx.node_spec
+    spec = ctx.agent_spec
    ctx_lines = [f"NODE: {spec.name} (id={spec.id})"]
    if spec.description:
        ctx_lines.append(f"PURPOSE: {spec.description}")
@@ -518,10 +554,7 @@ def build_llm_compaction_prompt(
        done = {k: v for k, v in acc.items() if v is not None}
        todo = [k for k, v in acc.items() if v is None]
        if done:
-            ctx_lines.append(
-                "OUTPUTS ALREADY SET:\n"
-                + "\n".join(f"  {k}: {str(v)[:150]}" for k, v in done.items())
-            )
+            ctx_lines.append("OUTPUTS ALREADY SET:\n" + "\n".join(f"  {k}: {str(v)[:150]}" for k, v in done.items()))
        if todo:
            ctx_lines.append(f"OUTPUTS STILL NEEDED: {', '.join(todo)}")
    elif spec.output_keys:
@@ -531,6 +564,18 @@ def build_llm_compaction_prompt(
    target_chars = target_tokens * 4
    node_ctx = "\n".join(ctx_lines)

+    user_messages_section = (
+        "6. **User Messages** — Reproduce EVERY user message verbatim and "
+        "in full, in chronological order, each on its own line prefixed "
+        'with the message index (e.g. "[U1] ..."). Do NOT paraphrase, '
+        "summarise, merge, or omit any user message. Preserve markdown, "
+        "code fences, whitespace, and punctuation exactly as the user "
+        "wrote them.\n"
+        if preserve_user_messages
+        else "6. **User Messages** — Preserve ALL user-stated rules, constraints, "
+        "identity preferences, and account details verbatim.\n"
+    )
+
    return (
        "You are compacting an AI agent's conversation history. "
        "The agent is still working and needs to continue.\n\n"
@@ -551,8 +596,7 @@ def build_llm_compaction_prompt(
        "resolved. Include root causes so the agent doesn't repeat them.\n"
        "5. **Problem Solving Efforts** — Approaches tried, dead ends hit, "
        "and reasoning behind the current strategy.\n"
-        "6. **User Messages** — Preserve ALL user-stated rules, constraints, "
-        "identity preferences, and account details verbatim.\n"
+        f"{user_messages_section}"
        "7. **Pending Tasks** — Work remaining, outputs still needed, and "
        "any blockers.\n"
        "8. **Current Work** — The most recent action taken and the immediate "
@@ -575,12 +619,8 @@ def build_message_inventory(conversation: NodeConversation) -> list[dict[str, An
        if message.tool_calls:
            for tool_call in message.tool_calls:
                args = tool_call.get("function", {}).get("arguments", "")
-                tool_call_args_chars += (
-                    len(args) if isinstance(args, str) else len(json.dumps(args))
-                )
-            names = [
-                tool_call.get("function", {}).get("name", "?") for tool_call in message.tool_calls
-            ]
+                tool_call_args_chars += len(args) if isinstance(args, str) else len(json.dumps(args))
+            names = [tool_call.get("function", {}).get("name", "?") for tool_call in message.tool_calls]
            tool_name = ", ".join(names)
        elif message.role == "tool" and message.tool_use_id:
            for previous in conversation.messages:
@@ -622,13 +662,13 @@ def write_compaction_debug_log(
    log_dir.mkdir(parents=True, exist_ok=True)

    ts = datetime.now(UTC).strftime("%Y%m%dT%H%M%S_%f")
-    node_label = ctx.node_id.replace("/", "_")
+    node_label = ctx.agent_id.replace("/", "_")
    log_path = log_dir / f"{ts}_{node_label}.md"

    lines: list[str] = [
-        f"# Compaction Debug — {ctx.node_id}",
+        f"# Compaction Debug — {ctx.agent_id}",
        f"**Time:** {datetime.now(UTC).isoformat()}",
-        f"**Node:** {ctx.node_spec.name} (`{ctx.node_id}`)",
+        f"**Node:** {ctx.agent_spec.name} (`{ctx.agent_id}`)",
    ]
    if ctx.stream_id:
        lines.append(f"**Stream:** {ctx.stream_id}")
@@ -637,14 +677,8 @@ def write_compaction_debug_log(
    lines.append("")

    if inventory:
-        total_chars = sum(
-            entry.get("content_chars", 0) + entry.get("tool_call_args_chars", 0)
-            for entry in inventory
-        )
-        lines.append(
-            "## Pre-Compaction Message Inventory "
-            f"({len(inventory)} messages, {total_chars:,} total chars)"
-        )
+        total_chars = sum(entry.get("content_chars", 0) + entry.get("tool_call_args_chars", 0) for entry in inventory)
+        lines.append(f"## Pre-Compaction Message Inventory ({len(inventory)} messages, {total_chars:,} total chars)")
        lines.append("")
        ranked = sorted(
            inventory,
@@ -663,8 +697,7 @@ def write_compaction_debug_log(
            if entry.get("phase"):
                flags.append(f"phase={entry['phase']}")
            lines.append(
-                f"| {i} | {entry['seq']} | {entry['role']} | {tool} "
-                f"| {chars:,} | {pct:.1f}% | {', '.join(flags)} |"
+                f"| {i} | {entry['seq']} | {entry['role']} | {tool} | {chars:,} | {pct:.1f}% | {', '.join(flags)} |"
            )

        large = [entry for entry in ranked if entry.get("preview")]
@@ -672,9 +705,7 @@ def write_compaction_debug_log(
            lines.append("")
            lines.append("### Large message previews")
            for entry in large:
-                lines.append(
-                    f"\n**seq={entry['seq']}** ({entry['role']}, {entry.get('tool', '')}):"
-                )
+                lines.append(f"\n**seq={entry['seq']}** ({entry['role']}, {entry.get('tool', '')}):")
                lines.append(f"```\n{entry['preview']}\n```")
    lines.append("")

@@ -715,7 +746,7 @@ async def log_compaction(

    if ctx.runtime_logger:
        ctx.runtime_logger.log_step(
-            node_id=ctx.node_id,
+            node_id=ctx.agent_id,
            node_type="event_loop",
            step_index=-1,
            llm_text=f"Context compacted ({level}): {before_pct}% \u2192 {after_pct}%",
@@ -724,7 +755,7 @@ async def log_compaction(
        )

    if event_bus:
-        from framework.runtime.event_bus import AgentEvent, EventType
+        from framework.host.event_bus import AgentEvent, EventType

        event_data: dict[str, Any] = {
            "level": level,
@@ -736,8 +767,8 @@ async def log_compaction(
        await event_bus.publish(
            AgentEvent(
                type=EventType.CONTEXT_COMPACTED,
-                stream_id=ctx.stream_id or ctx.node_id,
-                node_id=ctx.node_id,
+                stream_id=ctx.stream_id or ctx.agent_id,
+                node_id=ctx.agent_id,
                data=event_data,
            )
        )
@@ -762,13 +793,10 @@ def build_emergency_summary(
    node's known state so the LLM can continue working after
    compaction without losing track of its task and inputs.
    """
-    parts = [
-        "EMERGENCY COMPACTION — previous conversation was too large "
-        "and has been replaced with this summary.\n"
-    ]
+    parts = ["EMERGENCY COMPACTION — previous conversation was too large and has been replaced with this summary.\n"]

    # 1. Node identity
-    spec = ctx.node_spec
+    spec = ctx.agent_spec
    parts.append(f"NODE: {spec.name} (id={spec.id})")
    if spec.description:
        parts.append(f"PURPOSE: {spec.description}")
@@ -776,7 +804,7 @@ def build_emergency_summary(
    # 2. Inputs the node received
    input_lines = []
    for key in spec.input_keys:
-        value = ctx.input_data.get(key) or ctx.buffer.read(key)
+        value = ctx.input_data.get(key)
        if value is not None:
            # Truncate long values but keep them recognisable
            v_str = str(value)
@@ -818,28 +846,21 @@ def build_emergency_summary(
                data_files = [f for f in all_files if f not in conv_files]

                if conv_files:
-                    conv_list = "\n".join(
-                        f"  - {f}  (full path: {data_dir / f})" for f in conv_files
-                    )
+                    conv_list = "\n".join(f"  - {f}  (full path: {data_dir / f})" for f in conv_files)
                    parts.append(
                        "CONVERSATION HISTORY (freeform messages saved during compaction — "
-                        "use load_data('<filename>') to review earlier dialogue):\n" + conv_list
+                        "use read_file('<filename>') to review earlier dialogue):\n" + conv_list
                    )
                if data_files:
-                    file_list = "\n".join(
-                        f"  - {f}  (full path: {data_dir / f})" for f in data_files[:30]
-                    )
-                    parts.append("DATA FILES (use load_data('<filename>') to read):\n" + file_list)
+                    file_list = "\n".join(f"  - {f}  (full path: {data_dir / f})" for f in data_files[:30])
+                    parts.append("DATA FILES (use read_file('<filename>') to read):\n" + file_list)
                if not all_files:
                    parts.append(
                        "NOTE: Large tool results may have been saved to files. "
                        "Use list_directory to check the data directory."
                    )
        except Exception:
-            parts.append(
-                "NOTE: Large tool results were saved to files. "
-                "Use read_file(path='<path>') to read them."
-            )
+            parts.append("NOTE: Large tool results were saved to files. Use read_file(path='<path>') to read them.")

    # 6. Tool call history (prevent re-calling tools)
    if conversation is not None:
@@ -847,10 +868,7 @@ def build_emergency_summary(
        if tool_history:
            parts.append(tool_history)

-    parts.append(
-        "\nContinue working towards setting the remaining outputs. "
-        "Use your tools and the inputs above."
-    )
+    parts.append("\nContinue working towards setting the remaining outputs. Use your tools and the inputs above.")
    return "\n\n".join(parts)


@@ -861,6 +879,6 @@ def _extract_tool_call_history(conversation: NodeConversation) -> str:
    directly (vs. the module-level extract_tool_call_history in conversation.py
    which works on raw message lists).
    """
-    from framework.graph.conversation import extract_tool_call_history
+    from framework.agent_loop.conversation import extract_tool_call_history

    return extract_tool_call_history(list(conversation.messages))
@@ -12,12 +12,13 @@ import json
 import logging
 from collections.abc import Awaitable, Callable
 from dataclasses import dataclass
+from datetime import datetime
 from typing import Any

-from framework.graph.conversation import ConversationStore, NodeConversation
-from framework.graph.event_loop.types import LoopConfig, OutputAccumulator, TriggerEvent
-from framework.graph.node import NodeContext
+from framework.agent_loop.conversation import ConversationStore, NodeConversation
+from framework.agent_loop.internals.types import LoopConfig, OutputAccumulator, TriggerEvent
 from framework.llm.capabilities import supports_image_tool_results
+from framework.orchestrator.node import NodeContext

 logger = logging.getLogger(__name__)

@@ -53,15 +54,31 @@ async def restore(
    # continuous mode (or when _restore is called for timer-resume)
    # load all parts — the full conversation threads across nodes.
    _is_continuous = getattr(ctx, "continuous_mode", False)
-    phase_filter = None if _is_continuous else ctx.node_id
+    # The queen has agent_id="queen" but messages are stored with phase_id=None.
+    # Only apply phase filtering for non-queen workers in a multi-agent setup.
+    phase_filter = None if (_is_continuous or ctx.agent_id == "queen") else ctx.agent_id
    conversation = await NodeConversation.restore(
        conversation_store,
        phase_id=phase_filter,
        run_id=ctx.effective_run_id,
    )
    if conversation is None:
+        logger.info(
+            "[restore] No conversation found for agent_id=%s phase_filter=%s run_id=%s",
+            ctx.agent_id,
+            phase_filter,
+            ctx.effective_run_id,
+        )
        return None

+    logger.info(
+        "[restore] Restored %d messages for agent_id=%s phase_filter=%s run_id=%s",
+        conversation.message_count,
+        ctx.agent_id,
+        phase_filter,
+        ctx.effective_run_id,
+    )
+
    # If run_id filtering removed all messages, this is an intentional
    # restart (new run), not a crash recovery.  Return None so the caller
    # falls through to the fresh-conversation path.
@@ -124,7 +141,7 @@ async def write_cursor(
        cursor.update(
            {
                "iteration": iteration,
-                "node_id": ctx.node_id,
+                "node_id": ctx.agent_id,
                "outputs": accumulator.to_dict(),
            }
        )
@@ -133,9 +150,7 @@ async def write_cursor(
            cursor["recent_responses"] = recent_responses
        if recent_tool_fingerprints is not None:
            # Convert list[list[tuple]] → list[list[list]] for JSON
-            cursor["recent_tool_fingerprints"] = [
-                [list(pair) for pair in fps] for fps in recent_tool_fingerprints
-            ]
+            cursor["recent_tool_fingerprints"] = [[list(pair) for pair in fps] for fps in recent_tool_fingerprints]
        # Persist blocked-input state so restored runs re-block instead of
        # manufacturing a synthetic continuation turn.
        cursor["pending_input"] = pending_input
@@ -147,13 +162,14 @@ async def drain_injection_queue(
    conversation: NodeConversation,
    *,
    ctx: NodeContext,
-    describe_images_as_text_fn: (
-        Callable[[list[dict[str, Any]]], Awaitable[str | None]] | None
-    ) = None,
+    describe_images_as_text_fn: (Callable[[list[dict[str, Any]]], Awaitable[str | None]] | None) = None,
 ) -> int:
    """Drain all pending injected events as user messages. Returns count."""
    count = 0
-    logger.debug("[drain_injection_queue] Starting to drain queue, initial queue size: %s", queue.qsize() if hasattr(queue, 'qsize') else 'unknown')
+    logger.debug(
+        "[drain_injection_queue] Starting to drain queue, initial queue size: %s",
+        queue.qsize() if hasattr(queue, "qsize") else "unknown",
+    )
    while not queue.empty():
        try:
            content, is_client_input, image_content = queue.get_nowait()
@@ -176,15 +192,21 @@ async def drain_injection_queue(
                    else:
                        logger.info("[drain] no vision fallback available; images dropped")
                image_content = None
-            # Real user input is stored as-is; external events get a prefix
+            # Stamp every injected event with its arrival time so the model
+            # has a consistent temporal log to reason over (and so the
+            # stamp lives inside byte-stable conversation history instead
+            # of a per-turn system-prompt tail). Minute precision is what
+            # the queen needs for conversational / scheduling context.
+            stamp = datetime.now().astimezone().strftime("%Y-%m-%d %H:%M %Z")
            if is_client_input:
+                stamped = f"[{stamp}] {content}" if content else f"[{stamp}]"
                await conversation.add_user_message(
-                    content,
+                    stamped,
                    is_client_input=True,
                    image_content=image_content,
                )
            else:
-                await conversation.add_user_message(f"[External event]: {content}")
+                await conversation.add_user_message(f"[{stamp}] [External event] {content}")
            count += 1
        except asyncio.QueueEmpty:
            break
@@ -217,9 +239,12 @@ async def drain_trigger_queue(
        payload_str = json.dumps(t.payload, default=str)
        parts.append(f"[TRIGGER: {t.trigger_type}/{t.source_id}]{task_line}\n{payload_str}")

-    combined = "\n\n".join(parts)
+    stamp = datetime.now().astimezone().strftime("%Y-%m-%d %H:%M %Z")
+    combined = f"[{stamp}]\n" + "\n\n".join(parts)
    logger.info("[drain] %d trigger(s): %s", len(triggers), combined[:200])
-    await conversation.add_user_message(combined)
+    # Tag the message so the UI can render a banner instead of the raw
+    # `[TRIGGER: ...]` text. The LLM still sees `combined` verbatim.
+    await conversation.add_user_message(combined, is_trigger=True)
    return len(triggers)


@@ -242,11 +267,6 @@ async def check_pause(

    # Check context-level pause flags (legacy/alternative methods)
    pause_requested = ctx.input_data.get("pause_requested", False)
-    if not pause_requested:
-        try:
-            pause_requested = ctx.buffer.read("pause_requested") or False
-        except (PermissionError, KeyError):
-            pause_requested = False
    if pause_requested:
        completed = iteration
        logger.info(f"⏸ Pausing after {completed} iteration(s) completed (context-level)")
@@ -9,10 +9,10 @@ from __future__ import annotations
 import logging
 import time

-from framework.graph.conversation import NodeConversation
-from framework.graph.event_loop.types import HookContext
-from framework.graph.node import NodeContext
-from framework.runtime.event_bus import EventBus
+from framework.agent_loop.conversation import NodeConversation
+from framework.agent_loop.internals.types import HookContext
+from framework.host.event_bus import EventBus
+from framework.orchestrator.node import NodeContext

 logger = logging.getLogger(__name__)

@@ -45,14 +45,14 @@ async def generate_action_plan(
    Runs as a fire-and-forget task so it never blocks the main loop.
    """
    try:
-        system_prompt = ctx.node_spec.system_prompt or ""
+        system_prompt = ctx.agent_spec.system_prompt or ""
        # Trim to keep the prompt small
        prompt_summary = system_prompt[:500]
        if len(system_prompt) > 500:
            prompt_summary += "..."

        tool_names = [t.name for t in ctx.available_tools]
-        output_keys = ctx.node_spec.output_keys or []
+        output_keys = ctx.agent_spec.output_keys or []

        prompt = (
            f'You are about to work on a task as node "{node_id}".\n\n'
@@ -108,6 +108,8 @@ async def publish_llm_turn_complete(
    input_tokens: int,
    output_tokens: int,
    cached_tokens: int = 0,
+    cache_creation_tokens: int = 0,
+    cost_usd: float = 0.0,
    execution_id: str = "",
    iteration: int | None = None,
 ) -> None:
@@ -120,6 +122,8 @@ async def publish_llm_turn_complete(
            input_tokens=input_tokens,
            output_tokens=output_tokens,
            cached_tokens=cached_tokens,
+            cache_creation_tokens=cache_creation_tokens,
+            cost_usd=cost_usd,
            execution_id=execution_id,
            iteration=iteration,
        )
@@ -177,7 +181,7 @@ async def publish_context_usage(
    if not event_bus:
        return

-    from framework.runtime.event_bus import AgentEvent, EventType
+    from framework.host.event_bus import AgentEvent, EventType

    estimated = conversation.estimate_tokens()
    max_tokens = conversation._max_context_tokens
@@ -185,8 +189,8 @@ async def publish_context_usage(
    await event_bus.publish(
        AgentEvent(
            type=EventType.CONTEXT_USAGE_UPDATED,
-            stream_id=ctx.stream_id or ctx.node_id,
-            node_id=ctx.node_id,
+            stream_id=ctx.stream_id or ctx.agent_id,
+            node_id=ctx.agent_id,
            data={
                "usage_ratio": round(ratio, 4),
                "usage_pct": round(ratio * 100),
@@ -319,9 +323,7 @@ async def publish_output_key_set(
    execution_id: str = "",
 ) -> None:
    if event_bus:
-        await event_bus.emit_output_key_set(
-            stream_id=stream_id, node_id=node_id, key=key, execution_id=execution_id
-        )
+        pass


 async def run_hooks(
@@ -5,9 +5,9 @@ from __future__ import annotations
 import logging
 from collections.abc import Callable

-from framework.graph.conversation import NodeConversation
-from framework.graph.event_loop.types import JudgeProtocol, JudgeVerdict, OutputAccumulator
-from framework.graph.node import NodeContext
+from framework.agent_loop.conversation import NodeConversation
+from framework.agent_loop.internals.types import JudgeProtocol, JudgeVerdict, OutputAccumulator
+from framework.orchestrator.node import NodeContext

 logger = logging.getLogger(__name__)

@@ -31,14 +31,10 @@ class SubagentJudge:

        if remaining <= 3:
            urgency = (
-                f"URGENT: Only {remaining} iterations left. "
-                f"Stop all other work and call set_output NOW for: {missing}"
+                f"URGENT: Only {remaining} iterations left. Stop all other work and call set_output NOW for: {missing}"
            )
        elif remaining <= self._max_iterations // 2:
-            urgency = (
-                f"WARNING: {remaining} iterations remaining. "
-                f"You must call set_output for: {missing}"
-            )
+            urgency = f"WARNING: {remaining} iterations remaining. You must call set_output for: {missing}"
        else:
            urgency = f"Missing output keys: {missing}. Use set_output to provide them."

@@ -79,7 +75,7 @@ async def judge_turn(
    if mark_complete_flag:
        return JudgeVerdict(action="ACCEPT")

-    if ctx.node_spec.skip_judge:
+    if ctx.agent_spec.skip_judge:
        return JudgeVerdict(action="RETRY")  # feedback=None → not logged

    # --- Level 1: custom judge -----------------------------------------
@@ -92,9 +88,9 @@ async def judge_turn(
            "accumulator": accumulator,
            "iteration": iteration,
            "conversation_summary": conversation.export_summary(),
-            "output_keys": ctx.node_spec.output_keys,
+            "output_keys": ctx.agent_spec.output_keys,
            "missing_keys": get_missing_output_keys_fn(
-                accumulator, ctx.node_spec.output_keys, ctx.node_spec.nullable_output_keys
+                accumulator, ctx.agent_spec.output_keys, ctx.agent_spec.nullable_output_keys
            ),
        }
        verdict = await judge.evaluate(context)
@@ -109,9 +105,7 @@ async def judge_turn(
    if tool_results:
        return JudgeVerdict(action="RETRY")  # feedback=None → not logged

-    missing = get_missing_output_keys_fn(
-        accumulator, ctx.node_spec.output_keys, ctx.node_spec.nullable_output_keys
-    )
+    missing = get_missing_output_keys_fn(accumulator, ctx.agent_spec.output_keys, ctx.agent_spec.nullable_output_keys)

    if missing:
        return JudgeVerdict(
@@ -124,8 +118,8 @@ async def judge_turn(

    # All output keys present — run safety checks before accepting.

-    output_keys = ctx.node_spec.output_keys or []
-    nullable_keys = set(ctx.node_spec.nullable_output_keys or [])
+    output_keys = ctx.agent_spec.output_keys or []
+    nullable_keys = set(ctx.agent_spec.nullable_output_keys or [])

    # All-nullable with nothing set → node produced nothing useful.
    all_nullable = output_keys and nullable_keys >= set(output_keys)
@@ -133,36 +127,19 @@ async def judge_turn(
    if all_nullable and none_set:
        return JudgeVerdict(
            action="RETRY",
-            feedback=(
-                f"No output keys have been set yet. "
-                f"Use set_output to set at least one of: {output_keys}"
-            ),
-        )
-
-    # Queen with no output keys → continuous interaction node.
-    # Inject tool-use pressure instead of auto-accepting.
-    if not output_keys and ctx.supports_direct_user_io:
-        return JudgeVerdict(
-            action="RETRY",
-            feedback=(
-                "STOP describing what you will do. "
-                "You have FULL access to all tools — file creation, "
-                "shell commands, MCP tools — and you CAN call them "
-                "directly in your response. Respond ONLY with tool "
-                "calls, no prose. Execute the task now."
-            ),
+            feedback=(f"No output keys have been set yet. Use set_output to set at least one of: {output_keys}"),
        )

    # Level 2b: conversation-aware quality check (if success_criteria set)
-    if ctx.node_spec.success_criteria and ctx.llm:
-        from framework.graph.conversation_judge import evaluate_phase_completion
+    if ctx.agent_spec.success_criteria and ctx.llm:
+        from framework.orchestrator.conversation_judge import evaluate_phase_completion

        verdict = await evaluate_phase_completion(
            llm=ctx.llm,
            conversation=conversation,
-            phase_name=ctx.node_spec.name,
-            phase_description=ctx.node_spec.description,
-            success_criteria=ctx.node_spec.success_criteria,
+            phase_name=ctx.agent_spec.name,
+            phase_description=ctx.agent_spec.description,
+            success_criteria=ctx.agent_spec.success_criteria,
            accumulator_state=accumulator.to_dict(),
            max_context_tokens=max_context_tokens,
        )
@@ -0,0 +1,387 @@
+"""Synthetic tool builders for the event loop.
+
+Factory functions that create ``Tool`` definitions for framework-level
+synthetic tools (set_output, ask_user, escalate, delegate, report_to_parent).
+Also includes the ``handle_set_output`` validation logic.
+
+All functions are pure — they receive explicit parameters and return
+``Tool`` or ``ToolResult`` objects with no side effects.
+"""
+
+from __future__ import annotations
+
+from typing import Any
+
+from framework.llm.provider import Tool, ToolResult
+
+
+def sanitize_ask_user_inputs(
+    raw_question: Any,
+    raw_options: Any,
+) -> tuple[str, list[str] | None]:
+    """Self-heal a malformed ``ask_user`` tool call.
+
+    Some model families (notably when the system prompt teaches them
+    XML-ish scratchpad tags like ``<relationship>...</relationship>``)
+    carry that style into tool arguments and produce calls like::
+
+        ask_user({
+            "question": "What now?</question>\\n_OPTIONS: [\\"A\\", \\"B\\"]"
+        })
+
+    Symptoms:
+    - The chat UI renders ``</question>`` and ``_OPTIONS: [...]`` as
+      literal text in the question bubble.
+    - No buttons appear because the real ``options`` parameter is
+      empty.
+
+    This function:
+    - Strips leading/trailing whitespace.
+    - Removes a trailing ``</question>`` (with optional preceding
+      whitespace) from the question text.
+    - Detects an inline ``_OPTIONS:``, ``OPTIONS:``, or ``options:``
+      line followed by a JSON array, parses it, and returns the
+      recovered list as the second element.
+    - Removes the parsed line from the returned question text.
+
+    Returns ``(cleaned_question, recovered_options_or_None)``. The
+    caller should treat the recovered list as a fallback only when
+    the model did not also supply a real ``options`` array.
+    """
+    import json as _json
+    import re as _re
+
+    if raw_question is None:
+        return "", None
+    q = str(raw_question)
+
+    # Strip a stray </question> tag (case-insensitive, with optional
+    # preceding whitespace) anywhere in the string. This is the most
+    # common failure mode and never represents valid content.
+    q = _re.sub(r"\s*</\s*question\s*>\s*", "\n", q, flags=_re.IGNORECASE)
+
+    # Look for an inline options line. Match _OPTIONS, OPTIONS, options
+    # (with or without leading underscore), followed by ':' or '=', then
+    # a JSON array on the same line OR on the next line.
+    inline_options_re = _re.compile(
+        r"(?im)^\s*_?options\s*[:=]\s*(\[.*?\])\s*$",
+        _re.DOTALL,
+    )
+
+    recovered: list[str] | None = None
+    match = inline_options_re.search(q)
+    if match is not None:
+        try:
+            parsed = _json.loads(match.group(1))
+            if isinstance(parsed, list):
+                cleaned = [str(o).strip() for o in parsed if str(o).strip()]
+                if 1 <= len(cleaned) <= 8:
+                    recovered = cleaned
+        except (ValueError, TypeError):
+            pass
+        if recovered is not None:
+            # Remove the parsed line so it doesn't leak into the
+            # rendered question text.
+            q = inline_options_re.sub("", q, count=1)
+
+    # Strip any final whitespace / leftover blank lines from the
+    # question after removals.
+    q = _re.sub(r"\n{3,}", "\n\n", q).strip()
+
+    return q, recovered
+
+
+ask_user_prompt = """\
+Use this tool when you need to ask the user questions during execution. Reach for it when:
+
+- The task is ambiguous and the user needs to choose an approach
+- You need missing information to continue
+- You want approval before taking a meaningful action
+- A decision has real trade-offs the user should weigh in on
+- You want post-task feedback, or to offer saving a skill or updating memory
+
+Usage notes:
+- Users will always be able to select "Other" to provide custom text input, \
+so do not include catch-all options like "Other" or "Something else" yourself.
+- Each option is a plain string. Do NOT wrap options in `{"label": "..."}` or \
+`{"value": "..."}` objects — pass the raw choice text directly, e.g. `"Email"`, \
+not `{"label": "Email"}`.
+- If you recommend a specific option, make that the first option in the list \
+and append " (Recommended)" to the end of its text.
+- Call this tool whenever you need the user's response.
+- The prompt field must be plain text only.
+- Do not include XML, pseudo-tags, or inline option lists inside prompt.
+- Omit options only when the question truly requires a free-form response the \
+user must type out, such as describing an idea or pasting an error message.
+- Do not repeat the questions in your normal text response. The widget renders \
+them, so keep any surrounding text to a brief intro only.
+Example — single question with options:
+{"questions": [{"id": "next", "prompt": "What would you like to do?", \
+"options": ["Build a new agent (Recommended)", "Modify existing agent", "Run tests"]}]}
+
+Example — batch:
+{"questions": [
+  {"id": "scope", "prompt": "What scope?", "options": ["Full", "Partial"]},
+  {"id": "format", "prompt": "Output format?", "options": ["PDF", "CSV", "JSON"]},
+  {"id": "details", "prompt": "Any special requirements?"}
+]}
+
+Example — free-form (queen only):
+{"questions": [{"id": "idea", "prompt": "Describe the agent you want to build."}]}
+"""
+
+
+def build_ask_user_tool() -> Tool:
+    """Build the synthetic ask_user tool for explicit user-input requests.
+
+    The queen calls ask_user() when it needs to pause and wait for user
+    input. Accepts an array of 1-8 questions — a single question for the
+    common case, or a batch when several clarifications are needed at once.
+    Text-only turns WITHOUT ask_user flow through without blocking, allowing
+    progress updates and summaries to stream freely.
+    """
+    return Tool(
+        name="ask_user",
+        description=ask_user_prompt,
+        parameters={
+            "type": "object",
+            "properties": {
+                "questions": {
+                    "type": "array",
+                    "minItems": 1,
+                    "maxItems": 8,
+                    "description": "List of questions to present to the user.",
+                    "items": {
+                        "type": "object",
+                        "properties": {
+                            "id": {
+                                "type": "string",
+                                "description": ("Short identifier for this question (used in the response)."),
+                            },
+                            "prompt": {
+                                "type": "string",
+                                "description": "The question text shown to the user.",
+                            },
+                            "options": {
+                                "type": "array",
+                                "items": {"type": "string"},
+                                "description": (
+                                    "2-3 predefined choices as plain strings "
+                                    '(e.g. ["Yes", "No", "Maybe"]). Do NOT '
+                                    'wrap items in {"label": "..."} or '
+                                    '{"value": "..."} objects — pass the raw '
+                                    "choice text directly. The UI appends an "
+                                    "'Other' free-text input automatically, "
+                                    "so don't include catch-all options. "
+                                    "Omit only when the user must type a free-form answer."
+                                ),
+                                "minItems": 2,
+                                "maxItems": 3,
+                            },
+                        },
+                        "required": ["id", "prompt"],
+                    },
+                },
+            },
+            "required": ["questions"],
+        },
+    )
+
+
+def build_set_output_tool(output_keys: list[str] | None) -> Tool | None:
+    """Build the synthetic set_output tool for explicit output declaration."""
+    if not output_keys:
+        return None
+    return Tool(
+        name="set_output",
+        description=(
+            "Set an output value for this node. Call once per output key. "
+            "Use this for brief notes, counts, status, and file references — "
+            "NOT for large data payloads. When a tool result was saved to a "
+            "data file, pass the filename as the value "
+            "(e.g. 'google_sheets_get_values_1.txt') so the next phase can "
+            "load the full data. Values exceeding ~2000 characters are "
+            "auto-saved to data files. "
+            f"Valid keys: {output_keys}"
+        ),
+        parameters={
+            "type": "object",
+            "properties": {
+                "key": {
+                    "type": "string",
+                    "description": f"Output key. Must be one of: {output_keys}",
+                    "enum": output_keys,
+                },
+                "value": {
+                    "type": "string",
+                    "description": ("The output value — a brief note, count, status, or data filename reference."),
+                },
+            },
+            "required": ["key", "value"],
+        },
+    )
+
+
+def build_escalate_tool() -> Tool:
+    """Build the synthetic escalate tool for worker -> queen handoff."""
+    return Tool(
+        name="escalate",
+        description=(
+            "Escalate to the queen when requesting user input, "
+            "blocked by errors, missing "
+            "credentials, or ambiguous constraints that require supervisor "
+            "guidance. Include a concise reason and optional context. "
+            "The node will pause until the queen injects guidance."
+        ),
+        parameters={
+            "type": "object",
+            "properties": {
+                "reason": {
+                    "type": "string",
+                    "description": ("Short reason for escalation (e.g. 'Tool repeatedly failing')."),
+                },
+                "context": {
+                    "type": "string",
+                    "description": "Optional diagnostic details for the queen.",
+                },
+            },
+            "required": ["reason"],
+        },
+    )
+
+
+def build_report_to_parent_tool() -> Tool:
+    """Build the synthetic ``report_to_parent`` tool.
+
+    Parallel workers (those spawned by the overseer via
+    ``run_parallel_workers``) call this to send a structured report back
+    to the overseer queen when they have finished their task. Calling
+    ``report_to_parent`` terminates the worker's loop cleanly -- do not
+    call other tools after it.
+
+    The overseer receives these as ``SUBAGENT_REPORT`` events and
+    aggregates them into a single summary for the user.
+    """
+    return Tool(
+        name="report_to_parent",
+        description=(
+            "Send a structured report back to the parent overseer and "
+            "terminate. Call this when you have finished your task "
+            "(success, partial, or failed) or cannot make further "
+            "progress. Your loop ends after this call -- do not call any "
+            "other tool afterwards. The overseer reads the summary + "
+            "data fields and aggregates them into a user-facing response."
+        ),
+        parameters={
+            "type": "object",
+            "properties": {
+                "status": {
+                    "type": "string",
+                    "enum": ["success", "partial", "failed"],
+                    "description": (
+                        "Overall outcome. 'success' = task complete. "
+                        "'partial' = some progress but incomplete. "
+                        "'failed' = could not make progress."
+                    ),
+                },
+                "summary": {
+                    "type": "string",
+                    "description": (
+                        "One-paragraph narrative for the overseer. What "
+                        "you did, what you found, and any notable issues."
+                    ),
+                },
+                "data": {
+                    "type": "object",
+                    "description": (
+                        "Optional structured payload (rows fetched, IDs "
+                        "processed, files written, etc.) that the "
+                        "overseer can merge into its final summary."
+                    ),
+                },
+            },
+            "required": ["status", "summary"],
+        },
+    )
+
+
+def handle_report_to_parent(tool_input: dict[str, Any]) -> ToolResult:
+    """Normalise + validate a ``report_to_parent`` tool call.
+
+    Returns a ``ToolResult`` with the acknowledgement text the LLM sees;
+    the side effects (record on Worker, emit SUBAGENT_REPORT, terminate
+    loop) are performed by ``AgentLoop`` after this helper returns.
+    """
+    status = str(tool_input.get("status", "success")).strip().lower()
+    if status not in ("success", "partial", "failed"):
+        status = "success"
+    summary = str(tool_input.get("summary", "")).strip()
+    if not summary:
+        summary = f"(worker returned {status} with no summary)"
+    data = tool_input.get("data") or {}
+    if not isinstance(data, dict):
+        data = {"value": data}
+    # Store the normalised payload back on the input dict so the caller
+    # can pick it up without re-parsing.
+    tool_input["_normalised"] = {
+        "status": status,
+        "summary": summary,
+        "data": data,
+    }
+    return ToolResult(
+        tool_use_id=tool_input.get("tool_use_id", ""),
+        content=(f"Report delivered to overseer (status={status}). This worker will terminate now."),
+    )
+
+
+def handle_set_output(
+    tool_input: dict[str, Any],
+    output_keys: list[str] | None,
+) -> ToolResult:
+    """Handle set_output tool call. Returns ToolResult (sync)."""
+    import logging
+    import re
+
+    logger = logging.getLogger(__name__)
+
+    key = tool_input.get("key", "")
+    value = tool_input.get("value", "")
+    valid_keys = output_keys or []
+
+    # Recover from truncated JSON (max_tokens hit mid-argument).
+    # The _raw key is set by litellm when json.loads fails.
+    if not key and "_raw" in tool_input:
+        raw = tool_input["_raw"]
+        key_match = re.search(r'"key"\s*:\s*"(\w+)"', raw)
+        if key_match:
+            key = key_match.group(1)
+        val_match = re.search(r'"value"\s*:\s*"', raw)
+        if val_match:
+            start = val_match.end()
+            value = raw[start:].rstrip()
+            for suffix in ('"}\n', '"}', '"'):
+                if value.endswith(suffix):
+                    value = value[: -len(suffix)]
+                    break
+        if key:
+            logger.warning(
+                "Recovered set_output args from truncated JSON: key=%s, value_len=%d",
+                key,
+                len(value),
+            )
+            # Re-inject so the caller sees proper key/value
+            tool_input["key"] = key
+            tool_input["value"] = value
+
+    if key not in valid_keys:
+        return ToolResult(
+            tool_use_id="",
+            content=f"Invalid output key '{key}'. Valid keys: {valid_keys}",
+            is_error=True,
+        )
+
+    return ToolResult(
+        tool_use_id="",
+        content=f"Output '{key}' set successfully.",
+        is_error=False,
+    )
@@ -0,0 +1,291 @@
+"""Generic coercion of LLM-emitted tool arguments to match each tool's JSON schema.
+
+Small/mid-size models drift from tool schemas in predictable, boring ways:
+
+- A number field comes back as a string (``"42"`` instead of ``42``).
+- A boolean field comes back as a string (``"true"`` instead of ``True``).
+- An array-of-string field comes back as an array of objects
+  (``[{"label": "A"}, ...]`` instead of ``["A", ...]``).
+- An array/object field comes back as a JSON-encoded string
+  (``'["A","B"]'`` instead of ``["A", "B"]``).
+- A lone scalar arrives where the schema expects an array.
+
+This module centralizes the healing in one schema-driven pass that runs
+on every tool call before dispatch. Coercion is conservative:
+
+- Values that already match the expected type are untouched.
+- Shapes we don't recognize are returned as-is, so real bugs surface
+  instead of getting silently munged into something plausible.
+- Every actual coercion is logged with the tool, property, and shape
+  transition so we can see which models/tools are drifting.
+
+Tool-specific prompt drift (e.g. ``</question>`` tags leaking into an
+``ask_user`` prompt string) is NOT this module's job — that belongs in
+per-tool sanitizers, because it's about prompt style, not schema shape.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+from typing import Any
+
+from framework.llm.provider import Tool
+
+logger = logging.getLogger(__name__)
+
+# When an ``array<string>`` field arrives as an array of objects, look
+# for a text-carrying field in preference order. Covers the wrappers
+# small models tend to produce: ``[{"label": "A"}]``, ``[{"value": "A"}]``,
+# ``[{"text": "A"}]``, etc.
+_STRING_EXTRACT_KEYS: tuple[str, ...] = (
+    "label",
+    "value",
+    "text",
+    "name",
+    "title",
+    "display",
+)
+
+
+def coerce_tool_input(tool: Tool, raw_input: dict[str, Any] | None) -> dict[str, Any]:
+    """Coerce *raw_input* in place to match *tool*'s JSON schema.
+
+    Returns the mutated input dict (same object as *raw_input* when
+    possible, for callers that assume in-place mutation). Properties
+    not present in the schema are left untouched.
+    """
+    if not isinstance(raw_input, dict):
+        return raw_input or {}
+
+    schema = tool.parameters or {}
+    props = schema.get("properties")
+    if not isinstance(props, dict):
+        return raw_input
+
+    for key in list(raw_input.keys()):
+        prop_schema = props.get(key)
+        if not isinstance(prop_schema, dict):
+            continue
+        original = raw_input[key]
+        coerced = _coerce(original, prop_schema)
+        if coerced is not original:
+            logger.info(
+                "coerced tool input tool=%s prop=%s from=%s to=%s",
+                tool.name,
+                key,
+                _shape(original),
+                _shape(coerced),
+            )
+            raw_input[key] = coerced
+
+    return raw_input
+
+
+def _coerce(value: Any, schema: dict[str, Any]) -> Any:
+    """Dispatch on the schema's ``type`` field.
+
+    Returns the *same object* on passthrough so callers can detect
+    no-ops via identity (``coerced is value``).
+    """
+    expected = schema.get("type")
+    if not expected:
+        return value
+
+    # Union type: try each in order, return the first coercion that
+    # actually changes the value. Falls back to the original.
+    if isinstance(expected, list):
+        for t in expected:
+            sub_schema = {**schema, "type": t}
+            coerced = _coerce(value, sub_schema)
+            if coerced is not value:
+                return coerced
+        return value
+
+    if expected == "integer":
+        return _coerce_integer(value)
+    if expected == "number":
+        return _coerce_number(value)
+    if expected == "boolean":
+        return _coerce_boolean(value)
+    if expected == "string":
+        return _coerce_string(value)
+    if expected == "array":
+        return _coerce_array(value, schema)
+    if expected == "object":
+        return _coerce_object(value, schema)
+
+    return value
+
+
+def _coerce_integer(value: Any) -> Any:
+    # bool is a subclass of int in Python; don't mistake True for 1 here.
+    if isinstance(value, bool):
+        return value
+    if isinstance(value, int):
+        return value
+    if isinstance(value, str):
+        parsed = _parse_number(value)
+        if parsed is None:
+            return value
+        if parsed != int(parsed):
+            # Has a fractional part — caller asked for int, don't truncate.
+            return value
+        return int(parsed)
+    return value
+
+
+def _coerce_number(value: Any) -> Any:
+    if isinstance(value, bool):
+        return value
+    if isinstance(value, (int, float)):
+        return value
+    if isinstance(value, str):
+        parsed = _parse_number(value)
+        if parsed is None:
+            return value
+        if parsed == int(parsed):
+            return int(parsed)
+        return parsed
+    return value
+
+
+def _coerce_boolean(value: Any) -> Any:
+    if isinstance(value, bool):
+        return value
+    if isinstance(value, str):
+        low = value.strip().lower()
+        if low == "true":
+            return True
+        if low == "false":
+            return False
+    return value
+
+
+def _coerce_string(value: Any) -> Any:
+    if isinstance(value, str):
+        return value
+    # Common drift: model sent ``{"label": "..."}`` when we wanted "...".
+    if isinstance(value, dict):
+        extracted = _extract_string_from_object(value)
+        if extracted is not None:
+            return extracted
+    return value
+
+
+def _coerce_array(value: Any, schema: dict[str, Any]) -> Any:
+    # Heal: JSON-encoded array string → array.
+    if isinstance(value, str):
+        parsed = _try_parse_json(value)
+        if isinstance(parsed, list):
+            value = parsed
+        else:
+            # Scalar string where an array is expected — wrap it.
+            return [value]
+    elif not isinstance(value, list):
+        # Any other scalar (int, bool, dict, ...) — wrap.
+        return [value]
+
+    items_schema = schema.get("items")
+    if not isinstance(items_schema, dict):
+        return value
+
+    coerced_items: list[Any] = []
+    changed = False
+    for item in value:
+        c = _coerce(item, items_schema)
+        if c is not item:
+            changed = True
+        coerced_items.append(c)
+    return coerced_items if changed else value
+
+
+def _coerce_object(value: Any, schema: dict[str, Any]) -> Any:
+    # Heal: JSON-encoded object string → object.
+    if isinstance(value, str):
+        parsed = _try_parse_json(value)
+        if isinstance(parsed, dict):
+            value = parsed
+        else:
+            return value
+    if not isinstance(value, dict):
+        return value
+
+    sub_props = schema.get("properties")
+    if not isinstance(sub_props, dict):
+        return value
+
+    changed = False
+    for k in list(value.keys()):
+        sub_schema = sub_props.get(k)
+        if not isinstance(sub_schema, dict):
+            continue
+        original = value[k]
+        coerced = _coerce(original, sub_schema)
+        if coerced is not original:
+            value[k] = coerced
+            changed = True
+    # Return the same dict on mutation so callers that passed a shared
+    # reference see the updates. ``changed`` is only used to decide
+    # whether we need to log at a coarser level upstream.
+    return value if changed or not sub_props else value
+
+
+def _extract_string_from_object(obj: dict[str, Any]) -> str | None:
+    """Pick a likely-text field out of a wrapper object.
+
+    Tries the known keys first, falls back to the sole value if the
+    object has exactly one entry. Returns None when nothing plausible
+    is found — the caller keeps the original.
+    """
+    for k in _STRING_EXTRACT_KEYS:
+        v = obj.get(k)
+        if isinstance(v, str) and v:
+            return v
+    if len(obj) == 1:
+        (only,) = obj.values()
+        if isinstance(only, str) and only:
+            return only
+    return None
+
+
+def _try_parse_json(raw: str) -> Any:
+    try:
+        return json.loads(raw)
+    except (ValueError, TypeError):
+        return None
+
+
+def _parse_number(raw: str) -> float | None:
+    try:
+        f = float(raw)
+    except (ValueError, OverflowError):
+        return None
+    # Reject NaN and inf — they pass float() but aren't useful numeric
+    # values for tool arguments.
+    if f != f or f == float("inf") or f == float("-inf"):
+        return None
+    return f
+
+
+def _shape(value: Any) -> str:
+    """Short type/shape description used in coercion log lines."""
+    if value is None:
+        return "None"
+    if isinstance(value, bool):
+        return "bool"
+    if isinstance(value, int):
+        return "int"
+    if isinstance(value, float):
+        return "float"
+    if isinstance(value, str):
+        return f"str[{len(value)}]"
+    if isinstance(value, list):
+        if not value:
+            return "list[0]"
+        return f"list[{len(value)}]<{_shape(value[0])}>"
+    if isinstance(value, dict):
+        keys = sorted(value.keys())[:3]
+        suffix = ",…" if len(value) > 3 else ""
+        return f"dict{{{','.join(keys)}{suffix}}}"
+    return type(value).__name__
@@ -215,14 +215,30 @@ def truncate_tool_result(
    """Persist tool result to file and optionally truncate for context.

    When *spillover_dir* is configured, EVERY non-error tool result is
-    saved to a file (short filename like ``web_search_1.txt``).  A
-    ``[Saved to '...']`` annotation is appended so the reference
-    survives pruning and compaction.
+    written to disk for debugging. The LLM-visible content is then
+    shaped to avoid a **poison pattern** that we traced on 2026-04-15
+    through a gemini-3.1-pro-preview-customtools queen session: the prior format
+    appended ``\\n\\n[Saved to '/abs/path/file.txt']`` after every
+    small result, and frontier pattern-matching models (gemini 3.x in
+    particular) learned to autocomplete the `[Saved to '...']` trailer
+    in their own assistant turns, eventually degenerating into echoing
+    the whole tool result instead of deciding what to do next. See
+    ``session_20260415_100751_d49f4c28/conversations/parts/0000000056.json``
+    for the terminal case where the model's "text" output was the full
+    tool_result JSON.

-    - Small results (≤ limit): full content kept + file annotation
-    - Large results (> limit): preview + file reference
-    - Errors: pass through unchanged
-    - read_file/load_data results: truncate with pagination hint (no re-spill)
+    Rules after the fix:
+    - **Small results (≤ limit):** pass content through unchanged. No
+      trailer. No annotation. The full content is already in the
+      message; the disk copy is for debugging only.
+    - **Large results (> limit):** preview + file reference, but
+      formatted as plain prose instead of a bracketed ``[...]``
+      pattern. Structured JSON metadata ("_saved_to") is embedded
+      inside the JSON body when the preview is JSON-shaped so the
+      model can locate the full file without seeing a mimicry-prone
+      bracket token outside the body.
+    - **Errors:** pass through unchanged.
+    - **read_file results:** truncate with pagination hint (no re-spill).
    """
    limit = max_tool_result_chars

@@ -230,9 +246,9 @@ def truncate_tool_result(
    if result.is_error:
        return result

-    # read_file/load_data reads FROM spilled files — never re-spill (circular).
+    # read_file reads FROM spilled files — never re-spill (circular).
    # Just truncate with a pagination hint if the result is too large.
-    if tool_name in ("load_data", "read_file"):
+    if tool_name == "read_file":
        if limit <= 0 or len(result.content) <= limit:
            return result  # Small result — pass through as-is
        # Large result — truncate with smart preview
@@ -252,18 +268,19 @@ def truncate_tool_result(
        else:
            preview_block = result.content[:PREVIEW_CAP] + "…"

+        # Prose header (no brackets).
        header = (
-            f"[{tool_name} result: {len(result.content):,} chars — "
-            f"too large for context. Use offset_bytes/limit_bytes "
-            f"parameters to read smaller chunks.]"
+            f"Tool `{tool_name}` returned {len(result.content):,} characters "
+            f"(too large for context). Use offset_bytes / limit_bytes "
+            f"parameters to paginate smaller chunks."
        )
        if metadata_str:
            header += f"\n\nData structure:\n{metadata_str}"
        header += (
-            "\n\nWARNING: This is an INCOMPLETE preview. Do NOT draw conclusions or counts from it."
+            "\n\nWARNING: the preview below is a SAMPLE only — do NOT draw counts, totals, or conclusions from it."
        )

-        truncated = f"{header}\n\nPreview (small sample only):\n{preview_block}"
+        truncated = f"{header}\n\nPreview (truncated):\n{preview_block}"
        logger.info(
            "%s result truncated: %d → %d chars (use offset/limit to paginate)",
            tool_name,
@@ -301,7 +318,10 @@ def truncate_tool_result(

        if limit > 0 and len(result.content) > limit:
            # Large result: build a small, metadata-rich preview so the
-            # LLM cannot mistake it for the complete dataset.
+            # LLM cannot mistake it for the complete dataset. The
+            # preview is introduced as plain prose (no bracketed
+            # ``[Result from …]`` token) so it doesn't prime the model
+            # to autocomplete the same pattern in its next turn.
            PREVIEW_CAP = 5000

            # Extract structural metadata (array lengths, key names)
@@ -316,21 +336,21 @@ def truncate_tool_result(
            else:
                preview_block = result.content[:PREVIEW_CAP] + "…"

-            # Assemble header with structural info + warning
+            # Prose header (no brackets). Absolute path still surfaced
+            # so the agent can read the full file, but it's framed as
+            # a sentence, not a bracketed trailer.
            header = (
-                f"[Result from {tool_name}: {len(result.content):,} chars — "
-                f"too large for context, saved to '{abs_path}'.]\n"
+                f"Tool `{tool_name}` returned {len(result.content):,} characters "
+                f"(too large for context). Full result saved at: {abs_path}\n"
+                f"Read the complete data with read_file(path='{abs_path}').\n"
            )
            if metadata_str:
-                header += f"\nData structure:\n{metadata_str}"
+                header += f"\nData structure:\n{metadata_str}\n"
            header += (
-                f"\n\nWARNING: The preview below is INCOMPLETE. "
-                f"Do NOT draw conclusions or counts from it. "
-                f"Use read_file(path='{abs_path}') to read the "
-                f"full data before analysis."
+                "\nWARNING: the preview below is a SAMPLE only — do NOT draw counts, totals, or conclusions from it."
            )

-            content = f"{header}\n\nPreview (small sample only):\n{preview_block}"
+            content = f"{header}\n\nPreview (truncated):\n{preview_block}"
            logger.info(
                "Tool result spilled to file: %s (%d chars → %s)",
                tool_name,
@@ -338,10 +358,22 @@ def truncate_tool_result(
                abs_path,
            )
        else:
-            # Small result: keep full content + annotation with absolute path
-            content = f"{result.content}\n\n[Saved to '{abs_path}']"
+            # Small result: pass content through UNCHANGED.
+            #
+            # The prior design appended `\n\n[Saved to '/abs/path']`
+            # after every small result so the agent could re-read the
+            # file later. But (a) the full content is already in the
+            # message, so there's nothing to re-read; (b) the
+            # `[Saved to '…']` trailer is a repeating token pattern
+            # that frontier pattern-matching models autocomplete into
+            # their own assistant turns, eventually echoing whole tool
+            # results as "text" instead of making decisions. Dropping
+            # the trailer entirely kills the poison pattern. Spilled
+            # files on disk still exist for debugging — they just
+            # aren't advertised in the LLM-visible message.
+            content = result.content
            logger.info(
-                "Tool result saved to file: %s (%d chars → %s)",
+                "Tool result saved to file: %s (%d chars → %s, no trailer)",
                tool_name,
                len(result.content),
                filename,
@@ -373,15 +405,16 @@ def truncate_tool_result(
        else:
            preview_block = result.content[:PREVIEW_CAP] + "…"

+        # Prose header (no brackets) — see docstring for the poison
+        # pattern that the bracket format triggered.
        header = (
-            f"[Result from {tool_name}: {len(result.content):,} chars — "
-            f"truncated to fit context budget.]"
+            f"Tool `{tool_name}` returned {len(result.content):,} characters "
+            f"(truncated to fit context budget — no spillover dir configured)."
        )
        if metadata_str:
            header += f"\n\nData structure:\n{metadata_str}"
        header += (
-            "\n\nWARNING: This is an INCOMPLETE preview. "
-            "Do NOT draw conclusions or counts from the preview alone."
+            "\n\nWARNING: the preview below is a SAMPLE only — do NOT draw counts, totals, or conclusions from it."
        )

        truncated = f"{header}\n\n{preview_block}"
@@ -423,7 +456,7 @@ async def execute_tool(
        )

    skill_dirs = skill_dirs or []
-    skill_read_tools = {"view_file", "load_data", "read_file"}
+    skill_read_tools = {"view_file", "read_file"}
    if tc.tool_name in skill_read_tools and skill_dirs:
        raw_path = tc.tool_input.get("path", "")
        if raw_path:
@@ -467,6 +500,22 @@ async def execute_tool(
            result = await _run()
    except TimeoutError:
        logger.warning("Tool '%s' timed out after %.0fs", tc.tool_name, timeout)
+        # asyncio.wait_for cancels the awaiting coroutine, but the sync
+        # executor running inside run_in_executor keeps going — and so
+        # does any MCP subprocess it is blocked on. Reach through to the
+        # owning MCPClient and force-disconnect it so the subprocess is
+        # torn down. Next call_tool triggers a reconnect. Without this
+        # the executor thread and MCP child leak on every timeout.
+        kill_for_tool = getattr(tool_executor, "kill_for_tool", None)
+        if callable(kill_for_tool):
+            try:
+                await asyncio.to_thread(kill_for_tool, tc.tool_name)
+            except Exception as exc:  # defensive — never let cleanup crash the loop
+                logger.warning(
+                    "kill_for_tool('%s') raised during timeout handling: %s",
+                    tc.tool_name,
+                    exc,
+                )
        return ToolResult(
            tool_use_id=tc.tool_use_id,
            content=(
@@ -0,0 +1,323 @@
+"""Shared types and state containers for the event loop package."""
+
+from __future__ import annotations
+
+import asyncio
+import json
+import logging
+import time
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any, Literal, Protocol, runtime_checkable
+
+from framework.agent_loop.conversation import (
+    ConversationStore,
+)
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class TriggerEvent:
+    """A framework-level trigger signal (timer tick or webhook hit)."""
+
+    trigger_type: str
+    source_id: str
+    payload: dict[str, Any] = field(default_factory=dict)
+    timestamp: float = field(default_factory=time.time)
+
+
+@dataclass
+class JudgeVerdict:
+    """Result of judge evaluation for the event loop."""
+
+    action: Literal["ACCEPT", "RETRY", "ESCALATE"]
+    # None  = no evaluation happened (skip_judge, tool-continue); not logged.
+    # ""    = evaluated but no feedback; logged with default text.
+    # "..." = evaluated with feedback; logged as-is.
+    feedback: str | None = None
+
+
+@runtime_checkable
+class JudgeProtocol(Protocol):
+    """Protocol for event-loop judges."""
+
+    async def evaluate(self, context: dict[str, Any]) -> JudgeVerdict: ...
+
+
+@dataclass
+class LoopConfig:
+    """Configuration for the event loop."""
+
+    max_iterations: int = 50
+    # 0 (or any non-positive value) disables the per-turn hard limit,
+    # letting a single assistant turn fan out arbitrarily many tool
+    # calls. Models like Gemini 3.1 Pro routinely emit 40-80 tool
+    # calls in one turn during browser exploration; capping them
+    # strands work half-finished and makes the next turn repeat the
+    # discarded calls, which is worse than just running them.
+    max_tool_calls_per_turn: int = 0
+    judge_every_n_turns: int = 1
+    stall_detection_threshold: int = 3
+    stall_similarity_threshold: float = 0.85
+    max_context_tokens: int = 32_000
+    # Headroom reserved for the NEXT turn's input + output so that
+    # proactive compaction always finishes before the hard context limit
+    # is hit mid-stream. Scaled to match Claude Code's 13k-buffer-on-
+    # 200k-window ratio (~6.5%) applied to hive's default 32k window,
+    # with extra margin because hive's token estimator is char-based
+    # and less tight than Anthropic's own counting. Override via
+    # LoopConfig for larger windows.
+    compaction_buffer_tokens: int = 8_000
+    # Ratio-based component of the hybrid compaction buffer. Effective
+    # headroom reserved before compaction fires is
+    #   compaction_buffer_tokens + compaction_buffer_ratio * max_context_tokens
+    # The ratio scales with the model's window where the absolute fixed
+    # component does not (an 8k absolute buffer is 75% trigger on a 32k
+    # window but 96% on a 200k window). Combining them gives an absolute
+    # floor sized for the worst-case single tool result (one un-spilled
+    # max_tool_result_chars payload ≈ 30k chars ≈ 7.5k tokens, rounded to
+    # 8k) plus a fractional headroom that keeps the trigger meaningful on
+    # large windows, so the inner tool loop always has room to grow
+    # without tripping the mid-turn pre-send guard. Defaults: 8k + 15%.
+    # On 32k that's a 12.8k buffer (~60% trigger); on 200k it's 38k
+    # (~81% trigger); on 1M it's 158k (~84% trigger).
+    compaction_buffer_ratio: float = 0.15
+    # Warning is emitted one buffer earlier so the user/telemetry gets
+    # a "we're close" signal without triggering a compaction pass.
+    compaction_warning_buffer_tokens: int = 12_000
+    store_prefix: str = ""
+
+    # Overflow margin for max_tool_calls_per_turn. When the limit is
+    # enabled (>0), tool calls are only discarded when the count
+    # exceeds max_tool_calls_per_turn * (1 + margin). Ignored when
+    # max_tool_calls_per_turn is 0.
+    tool_call_overflow_margin: float = 0.5
+
+    # Tool result context management.
+    max_tool_result_chars: int = 30_000
+    spillover_dir: str | None = None
+
+    # Image retention in conversation history.
+    # Screenshots from ``browser_screenshot`` are inlined as base64
+    # data URLs inside message ``image_content``. Each full-page
+    # screenshot costs ~250k tokens when the provider counts the
+    # base64 as text (gemini, most non-Anthropic providers). Four
+    # screenshots in one conversation push gemini's 1M context over
+    # the limit and the model starts emitting garbage.
+    #
+    # The framework strips image_content from older messages after
+    # every tool-result batch, keeping only the most recent N
+    # screenshots. The text metadata on evicted messages (url, size,
+    # scale hints) is preserved so the agent can still reason about
+    # "I took a screenshot at step N that showed the compose modal".
+    # Raise this only if you genuinely need longer visual history AND
+    # you know your provider is using native image tokenization.
+    max_retained_screenshots: int = 2
+
+    # set_output value spilling.
+    max_output_value_chars: int = 2_000
+
+    # Stream retry.
+    max_stream_retries: int = 5
+    stream_retry_backoff_base: float = 2.0
+    stream_retry_max_delay: float = 60.0
+    # Persistent retry for capacity-class errors (429, 529, overloaded).
+    # Unlike the bounded retry above, these keep trying until the wall-clock
+    # budget below is exhausted — modelled after claude-code's withRetry.
+    # The loop still publishes a retry event each attempt so the UI can
+    # see progress. Set to 0 to disable and fall back to bounded retry.
+    capacity_retry_max_seconds: float = 600.0
+    capacity_retry_max_delay: float = 60.0
+
+    # Tool doom loop detection.
+    tool_doom_loop_threshold: int = 3
+
+    # Client-facing auto-block grace period.
+    cf_grace_turns: int = 1
+    # Worker auto-escalation: text-only turns before escalating to queen.
+    worker_escalation_grace_turns: int = 1
+    tool_doom_loop_enabled: bool = True
+    # Silent worker: consecutive tool-only turns (no user-facing text)
+    # before injecting a nudge to communicate progress.
+    silent_tool_streak_threshold: int = 5
+
+    # Per-tool-call timeout.
+    tool_call_timeout_seconds: float = 60.0
+
+    # LLM stream inactivity watchdog. Split into two budgets so legitimate
+    # slow TTFT on large contexts doesn't get mistaken for a dead connection.
+    # - ttft: stream open -> first event. Large-context local models can
+    #   legitimately take minutes before the first token arrives.
+    # - inter_event: last event -> now, ONLY after the first event. A stream
+    #   that started producing and then went silent is a real stall.
+    # Whichever fires first cancels the stream. Set to 0 to disable that
+    # individual budget; set both to 0 to fully disable the watchdog.
+    llm_stream_ttft_timeout_seconds: float = 600.0
+    llm_stream_inter_event_idle_seconds: float = 120.0
+    # Deprecated alias — kept so existing configs keep working. If set to a
+    # non-default value it overrides inter_event_idle (historical behavior).
+    llm_stream_inactivity_timeout_seconds: float = 120.0
+
+    # Continue-nudge recovery. When the idle watchdog fires on a live but
+    # stuck stream, cancel the stream and append a short continuation
+    # hint to the conversation instead of raising a ConnectionError and
+    # re-running the whole turn. Preserves any partial text/tool-calls the
+    # stream emitted before the stall.
+    continue_nudge_enabled: bool = True
+    # Cap so a truly dead endpoint eventually falls back to the error path
+    # instead of nudging forever.
+    continue_nudge_max_per_turn: int = 3
+
+    # Tool-call replay detector. When the model emits a tool call whose
+    # (name + canonical-args) matches a prior successful call in the last
+    # K assistant turns, emit telemetry and prepend a short steer onto the
+    # tool result — but still execute. Weaker models legitimately repeat
+    # read-only calls (screenshot, evaluate), so silent skipping would
+    # cause surprising behavior.
+    replay_detector_enabled: bool = True
+    replay_detector_within_last_turns: int = 3
+
+    # Subagent delegation timeout (wall-clock max).
+    subagent_timeout_seconds: float = 3600.0
+
+    # Subagent inactivity timeout - only timeout if no activity for this duration.
+    # This resets whenever the subagent makes progress (tool calls, LLM responses).
+    # Set to 0 to use only the wall-clock timeout.
+    subagent_inactivity_timeout_seconds: float = 300.0
+
+    # Lifecycle hooks.
+    hooks: dict[str, list] | None = None
+
+    def __post_init__(self) -> None:
+        if self.hooks is None:
+            object.__setattr__(self, "hooks", {})
+
+
+@dataclass
+class HookContext:
+    """Context passed to every lifecycle hook."""
+
+    event: str
+    trigger: str | None
+    system_prompt: str
+
+
+@dataclass
+class HookResult:
+    """What a hook may return to modify node state."""
+
+    system_prompt: str | None = None
+    inject: str | None = None
+
+
+@dataclass
+class OutputAccumulator:
+    """Accumulates output key-value pairs with optional write-through persistence."""
+
+    values: dict[str, Any] = field(default_factory=dict)
+    store: ConversationStore | None = None
+    spillover_dir: str | None = None
+    max_value_chars: int = 0
+    run_id: str | None = None
+
+    async def set(self, key: str, value: Any) -> None:
+        """Set a key-value pair, auto-spilling large values to files."""
+        value = await self._auto_spill(key, value)
+        self.values[key] = value
+        if self.store:
+            cursor = await self.store.read_cursor() or {}
+            outputs = cursor.get("outputs", {})
+            outputs[key] = value
+            cursor["outputs"] = outputs
+            await self.store.write_cursor(cursor)
+
+    async def _auto_spill(self, key: str, value: Any) -> Any:
+        """Save large values to a file and return a reference string.
+
+        Runs the JSON serialization and file write on a worker thread
+        so they don't block the asyncio event loop. For a 100k-char
+        dict this used to freeze every concurrent tool call for ~50ms
+        of ``json.dumps(indent=2)`` + a sync disk write; for bigger
+        payloads or slow storage (NFS, networked FS) the freeze was
+        proportionally worse.
+        """
+        if self.max_value_chars <= 0 or not self.spillover_dir:
+            return value
+
+        # Cheap size probe first — if the value is already a short
+        # string we can skip both the JSON round-trip and the thread
+        # hop entirely.
+        if isinstance(value, str) and len(value) <= self.max_value_chars:
+            return value
+
+        def _spill_sync() -> Any:
+            # JSON serialization for size check (only for non-strings).
+            if isinstance(value, str):
+                val_str = value
+            else:
+                val_str = json.dumps(value, ensure_ascii=False)
+            if len(val_str) <= self.max_value_chars:
+                return value
+
+            spill_path = Path(self.spillover_dir)
+            spill_path.mkdir(parents=True, exist_ok=True)
+            ext = ".json" if isinstance(value, (dict, list)) else ".txt"
+            filename = f"output_{key}{ext}"
+            write_content = (
+                json.dumps(value, indent=2, ensure_ascii=False) if isinstance(value, (dict, list)) else str(value)
+            )
+            file_path = spill_path / filename
+            file_path.write_text(write_content, encoding="utf-8")
+            file_size = file_path.stat().st_size
+            logger.info(
+                "set_output value auto-spilled: key=%s, %d chars -> %s (%d bytes)",
+                key,
+                len(val_str),
+                filename,
+                file_size,
+            )
+            # Use absolute path so parent agents can find files from subagents.
+            #
+            # Prose format (no brackets) — same fix as tool_result_handler:
+            # frontier pattern-matching models autocomplete bracketed
+            # `[Saved to '...']` trailers into their own assistant turns,
+            # eventually degenerating into echoing the file path as text.
+            # Keep the path accessible but frame it as plain prose.
+            abs_path = str(file_path.resolve())
+            return (
+                f"Output saved at: {abs_path} ({file_size:,} bytes). "
+                f"Read the full data with read_file(path='{abs_path}')."
+            )
+
+        return await asyncio.to_thread(_spill_sync)
+
+    def get(self, key: str) -> Any | None:
+        return self.values.get(key)
+
+    def to_dict(self) -> dict[str, Any]:
+        return dict(self.values)
+
+    def has_all_keys(self, required: list[str]) -> bool:
+        return all(key in self.values and self.values[key] is not None for key in required)
+
+    @classmethod
+    async def restore(
+        cls,
+        store: ConversationStore,
+        run_id: str | None = None,
+    ) -> OutputAccumulator:
+        cursor = await store.read_cursor()
+        values = cursor.get("outputs", {}) if cursor else {}
+        return cls(values=values, store=store, run_id=run_id)
+
+
+__all__ = [
+    "HookContext",
+    "HookResult",
+    "JudgeProtocol",
+    "JudgeVerdict",
+    "LoopConfig",
+    "OutputAccumulator",
+    "TriggerEvent",
+]
@@ -0,0 +1,105 @@
+"""Prompt composition for agent loops.
+
+Builds canonical system prompts from AgentContext fields.
+Extracted from the former orchestrator/prompting module.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from datetime import datetime
+from typing import Any
+
+
+@dataclass(frozen=True)
+class PromptSpec:
+    identity_prompt: str = ""
+    focus_prompt: str = ""
+    narrative: str = ""
+    accounts_prompt: str = ""
+    skills_catalog_prompt: str = ""
+    protocols_prompt: str = ""
+    memory_prompt: str = ""
+    agent_type: str = "event_loop"
+    output_keys: tuple[str, ...] = ()
+
+
+def stamp_prompt_datetime(prompt: str) -> str:
+    local = datetime.now().astimezone()
+    stamp = f"Current date and time: {local.strftime('%Y-%m-%d %H:%M %Z (UTC%z)')}"
+    return f"{prompt}\n\n{stamp}" if prompt else stamp
+
+
+def build_prompt_spec(
+    ctx: Any,
+    *,
+    focus_prompt: str | None = None,
+    narrative: str | None = None,
+    memory_prompt: str | None = None,
+) -> PromptSpec:
+    from framework.skills.tool_gating import augment_catalog_for_tools
+
+    resolved_memory = memory_prompt
+    if resolved_memory is None:
+        resolved_memory = getattr(ctx, "memory_prompt", "") or ""
+        dynamic = getattr(ctx, "dynamic_memory_provider", None)
+        if dynamic is not None:
+            try:
+                resolved_memory = dynamic() or ""
+            except Exception:
+                resolved_memory = getattr(ctx, "memory_prompt", "") or ""
+
+    # Tool-gated pre-activation: inject full body of default skills whose
+    # trigger tools are present in this agent's tool list (e.g. browser_*
+    # pulls in hive.browser-automation). Keeps non-browser agents lean.
+    tool_names = [getattr(t, "name", "") for t in (getattr(ctx, "available_tools", None) or [])]
+    raw_catalog = ctx.skills_catalog_prompt or ""
+    dynamic_catalog = getattr(ctx, "dynamic_skills_catalog_provider", None)
+    if dynamic_catalog is not None:
+        try:
+            raw_catalog = dynamic_catalog() or ""
+        except Exception:
+            raw_catalog = ctx.skills_catalog_prompt or ""
+    skills_catalog_prompt = augment_catalog_for_tools(raw_catalog, tool_names)
+
+    return PromptSpec(
+        identity_prompt=ctx.identity_prompt or "",
+        focus_prompt=focus_prompt if focus_prompt is not None else (ctx.agent_spec.system_prompt or ""),
+        narrative=narrative if narrative is not None else (ctx.narrative or ""),
+        accounts_prompt=ctx.accounts_prompt or "",
+        skills_catalog_prompt=skills_catalog_prompt,
+        protocols_prompt=ctx.protocols_prompt or "",
+        memory_prompt=resolved_memory,
+        agent_type=ctx.agent_spec.agent_type,
+        output_keys=tuple(ctx.agent_spec.output_keys or ()),
+    )
+
+
+def build_system_prompt(spec: PromptSpec) -> str:
+    parts: list[str] = []
+    if spec.identity_prompt:
+        parts.append(spec.identity_prompt)
+    if spec.accounts_prompt:
+        parts.append(f"\n{spec.accounts_prompt}")
+    if spec.skills_catalog_prompt:
+        parts.append(f"\n{spec.skills_catalog_prompt}")
+    if spec.protocols_prompt:
+        parts.append(f"\n{spec.protocols_prompt}")
+    if spec.memory_prompt:
+        parts.append(f"\n{spec.memory_prompt}")
+    if spec.focus_prompt:
+        parts.append(f"\n{spec.focus_prompt}")
+    if spec.narrative:
+        parts.append(f"\n{spec.narrative}")
+    return "\n".join(parts)
+
+
+def build_system_prompt_for_context(
+    ctx: Any,
+    *,
+    focus_prompt: str | None = None,
+    narrative: str | None = None,
+    memory_prompt: str | None = None,
+) -> str:
+    spec = build_prompt_spec(ctx, focus_prompt=focus_prompt, narrative=narrative, memory_prompt=memory_prompt)
+    return build_system_prompt(spec)
@@ -0,0 +1,281 @@
+"""Core types for the agent loop — the execution primitive of the colony.
+
+AgentSpec:    Declarative definition of what an agent does.
+AgentContext: Everything an agent loop needs to execute.
+AgentResult:  What comes out of an agent loop execution.
+AgentProtocol: Interface that all agent implementations must satisfy.
+"""
+
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+from dataclasses import dataclass, field
+from typing import Any
+
+from pydantic import BaseModel, Field
+
+from framework.llm.provider import LLMProvider, Tool
+from framework.tracker.decision_tracker import DecisionTracker
+
+
+class AgentSpec(BaseModel):
+    """Declarative definition of an agent's capabilities and configuration.
+
+    This is the blueprint from which AgentLoop instances are created.
+    Workers in a colony are exact copies of the queen's AgentSpec.
+    """
+
+    id: str
+    name: str
+    description: str
+
+    agent_type: str = Field(
+        default="event_loop",
+        description="Type: 'event_loop' (recommended), 'gcu' (browser automation).",
+    )
+
+    input_keys: list[str] = Field(
+        default_factory=list,
+        description="Keys this agent reads from input data",
+    )
+    output_keys: list[str] = Field(
+        default_factory=list,
+        description="Keys this agent produces as output",
+    )
+    nullable_output_keys: list[str] = Field(
+        default_factory=list,
+        description="Output keys that can be None without triggering validation errors",
+    )
+
+    input_schema: dict[str, dict] = Field(
+        default_factory=dict,
+        description="Optional schema for input validation.",
+    )
+    output_schema: dict[str, dict] = Field(
+        default_factory=dict,
+        description="Optional schema for output validation.",
+    )
+
+    system_prompt: str | None = Field(default=None, description="System prompt for the LLM")
+    tools: list[str] = Field(default_factory=list, description="Tool names this agent can use")
+    tool_access_policy: str = Field(
+        default="explicit",
+        description=(
+            "'all' = all tools from registry, "
+            "'explicit' = only tools listed in `tools` (default), "
+            "'none' = no tools at all."
+        ),
+    )
+    model: str | None = Field(default=None, description="Specific model override")
+
+    function: str | None = Field(default=None, description="Function name or path")
+    routes: dict[str, str] = Field(default_factory=dict, description="Condition -> target mapping")
+
+    max_retries: int = Field(default=3)
+    retry_on: list[str] = Field(default_factory=list, description="Error types to retry on")
+
+    max_visits: int = Field(
+        default=0,
+        description=("Max times this agent executes in one colony run. 0 = unlimited. Set >1 for one-shot agents."),
+    )
+
+    output_model: type[BaseModel] | None = Field(
+        default=None,
+        description="Optional Pydantic model for validating LLM output.",
+    )
+    max_validation_retries: int = Field(
+        default=2,
+        description="Maximum retries when Pydantic validation fails",
+    )
+
+    client_facing: bool = Field(
+        default=False,
+        description="Deprecated — the queen is intrinsically interactive.",
+    )
+
+    success_criteria: str | None = Field(
+        default=None,
+        description="Natural-language criteria for phase completion.",
+    )
+
+    skip_judge: bool = Field(
+        default=False,
+        description="When True, the implicit judge is bypassed entirely.",
+    )
+
+    model_config = {"extra": "allow", "arbitrary_types_allowed": True}
+
+    def is_queen(self) -> bool:
+        return self.id == "queen"
+
+    def supports_direct_user_io(self) -> bool:
+        return self.is_queen()
+
+
+def deprecated_client_facing_warning(spec: AgentSpec) -> str | None:
+    if spec.client_facing and not spec.is_queen():
+        return (
+            f"Agent '{spec.id}' sets deprecated client_facing=True. "
+            "Non-queen direct human I/O is no longer supported; route worker "
+            "questions and approvals through queen escalation instead."
+        )
+    return None
+
+
+def warn_if_deprecated_client_facing(spec: AgentSpec) -> None:
+    import logging
+
+    warning = deprecated_client_facing_warning(spec)
+    if warning:
+        logging.getLogger(__name__).warning(warning)
+
+
+@dataclass
+class AgentContext:
+    """Everything an agent loop needs to execute.
+
+    Passed to every agent implementation and provides:
+    - Runtime (for decision logging)
+    - LLM access
+    - Tools
+    - Goal context
+    - Execution metadata
+    """
+
+    runtime: DecisionTracker
+
+    agent_id: str
+    agent_spec: AgentSpec
+
+    input_data: dict[str, Any] = field(default_factory=dict)
+
+    llm: LLMProvider | None = None
+    available_tools: list[Tool] = field(default_factory=list)
+
+    goal_context: str = ""
+    goal: Any = None
+
+    max_tokens: int = 4096
+
+    attempt: int = 1
+    max_attempts: int = 3
+
+    runtime_logger: Any = None
+    pause_event: Any = None
+
+    accounts_prompt: str = ""
+
+    identity_prompt: str = ""
+    narrative: str = ""
+    memory_prompt: str = ""
+
+    event_triggered: bool = False
+
+    execution_id: str = ""
+    run_id: str = ""
+
+    @property
+    def effective_run_id(self) -> str | None:
+        return self.run_id or None
+
+    stream_id: str = ""
+
+    dynamic_tools_provider: Any = None
+    dynamic_prompt_provider: Any = None
+    # Optional Callable[[], str]: when set alongside ``dynamic_prompt_provider``,
+    # the AgentLoop sends the system prompt as two pieces — the result of
+    # ``dynamic_prompt_provider`` is the STATIC block (cached), and this
+    # provider returns the DYNAMIC suffix (not cached). The LLM wrapper
+    # emits them as two Anthropic system content blocks with a cache
+    # breakpoint between them for providers that honor ``cache_control``.
+    # For providers that don't, the two strings are concatenated. Used by
+    # the Queen to keep her persona/role/tools block warm across iterations
+    # while the recall + timestamp tail refreshes per user turn.
+    dynamic_prompt_suffix_provider: Any = None
+    dynamic_memory_provider: Any = None
+    # Optional Callable[[], str]: when set, the current skills-catalog
+    # prompt is sourced from this provider each iteration. Lets workers
+    # pick up UI toggles without restarting the run. Queen agents already
+    # rebuild the whole prompt via dynamic_prompt_provider — this field
+    # is a surgical alternative used by colony workers where the rest of
+    # the prompt stays constant and we don't want to thrash the cache.
+    dynamic_skills_catalog_provider: Any = None
+
+    skills_catalog_prompt: str = ""
+    protocols_prompt: str = ""
+    skill_dirs: list[str] = field(default_factory=list)
+    default_skill_batch_nudge: str | None = None
+    default_skill_warn_ratio: float | None = None
+
+    iteration_metadata_provider: Any = None
+
+    @property
+    def is_queen_stream(self) -> bool:
+        return self.stream_id == "queen" or self.agent_spec.is_queen()
+
+    @property
+    def emits_client_io(self) -> bool:
+        return self.is_queen_stream
+
+    @property
+    def supports_direct_user_io(self) -> bool:
+        return self.is_queen_stream and not self.event_triggered
+
+
+@dataclass
+class AgentResult:
+    """Output of an agent loop execution."""
+
+    success: bool
+    output: dict[str, Any] = field(default_factory=dict)
+    error: str | None = None
+
+    next_agent: str | None = None
+    route_reason: str | None = None
+
+    tokens_used: int = 0
+    latency_ms: int = 0
+
+    validation_errors: list[str] = field(default_factory=list)
+
+    conversation: Any = None
+
+    # Machine-readable reason the loop stopped (see LoopExitReason in
+    # agent_loop/internals/types.py). "?" means the loop didn't set one,
+    # which should itself be treated as a diagnostic.
+    exit_reason: str = "?"
+    # Counters for reliability events surfaced during this execution.
+    # Populated from the loop's TaskRegistry-style counters at return
+    # time so callers can spot recurring failure modes without tailing
+    # logs. Keys are stable strings; missing keys mean "zero".
+    reliability_stats: dict[str, int] = field(default_factory=dict)
+
+    def to_summary(self, spec: Any = None) -> str:
+        if not self.success:
+            return f"Failed: {self.error}"
+
+        if not self.output:
+            return "Completed (no output)"
+
+        parts = [f"Completed with {len(self.output)} outputs:"]
+        for key, value in list(self.output.items())[:5]:
+            value_str = str(value)[:100]
+            if len(str(value)) > 100:
+                value_str += "..."
+            parts.append(f"  - {key}: {value_str}")
+        return "\n".join(parts)
+
+
+class AgentProtocol(ABC):
+    """Interface all agent implementations must satisfy."""
+
+    @abstractmethod
+    async def execute(self, ctx: AgentContext) -> AgentResult:
+        pass
+
+    def validate_input(self, ctx: AgentContext) -> list[str]:
+        errors = []
+        for key in ctx.agent_spec.input_keys:
+            if key not in ctx.input_data:
+                errors.append(f"Missing required input: {key}")
+        return errors
@@ -8,6 +8,10 @@ FRAMEWORK_AGENTS_DIR = Path(__file__).parent
 def list_framework_agents() -> list[Path]:
    """List all framework agent directories."""
    return sorted(
-        [p for p in FRAMEWORK_AGENTS_DIR.iterdir() if p.is_dir() and (p / "agent.py").exists()],
+        [
+            p
+            for p in FRAMEWORK_AGENTS_DIR.iterdir()
+            if p.is_dir() and ((p / "agent.json").exists() or (p / "agent.py").exists())
+        ],
        key=lambda p: p.name,
    )
@@ -21,15 +21,15 @@ from pathlib import Path
 from typing import TYPE_CHECKING

 from framework.config import get_max_context_tokens
-from framework.graph import Goal, NodeSpec, SuccessCriterion
-from framework.graph.checkpoint_config import CheckpointConfig
-from framework.graph.edge import GraphSpec
-from framework.graph.executor import ExecutionResult
+from framework.host.agent_host import AgentHost
+from framework.host.execution_manager import EntryPointSpec
 from framework.llm import LiteLLMProvider
-from framework.runner.mcp_registry import MCPRegistry
-from framework.runner.tool_registry import ToolRegistry
-from framework.runtime.agent_runtime import AgentRuntime, create_agent_runtime
-from framework.runtime.execution_stream import EntryPointSpec
+from framework.loader.mcp_registry import MCPRegistry
+from framework.loader.tool_registry import ToolRegistry
+from framework.orchestrator import Goal, NodeSpec, SuccessCriterion
+from framework.orchestrator.checkpoint_config import CheckpointConfig
+from framework.orchestrator.edge import GraphSpec
+from framework.orchestrator.orchestrator import ExecutionResult

 from .config import default_config
 from .nodes import build_tester_node
@@ -37,7 +37,7 @@ from .nodes import build_tester_node
 logger = logging.getLogger(__name__)

 if TYPE_CHECKING:
-    from framework.runner import AgentRunner
+    from framework.loader import AgentLoader

 logger = logging.getLogger(__name__)

@@ -126,9 +126,7 @@ def _list_local_accounts() -> list[dict]:
    try:
        from framework.credentials.local.registry import LocalCredentialRegistry

-        return [
-            info.to_account_dict() for info in LocalCredentialRegistry.default().list_accounts()
-        ]
+        return [info.to_account_dict() for info in LocalCredentialRegistry.default().list_accounts()]
    except ImportError as exc:
        logger.debug("Local credential registry unavailable: %s", exc)
        return []
@@ -181,9 +179,7 @@ def _list_env_fallback_accounts() -> list[dict]:
            if spec.credential_group in seen_groups:
                continue
            group_available = all(
-                _is_configured(n, s)
-                for n, s in CREDENTIAL_SPECS.items()
-                if s.credential_group == spec.credential_group
+                _is_configured(n, s) for n, s in CREDENTIAL_SPECS.items() if s.credential_group == spec.credential_group
            )
            if not group_available:
                continue
@@ -215,9 +211,7 @@ def list_connected_accounts() -> list[dict]:

    # Show env-var fallbacks only for credentials not already in the named registry
    local_providers = {a["provider"] for a in local}
-    env_fallbacks = [
-        a for a in _list_env_fallback_accounts() if a["provider"] not in local_providers
-    ]
+    env_fallbacks = [a for a in _list_env_fallback_accounts() if a["provider"] not in local_providers]

    return aden + local + env_fallbacks

@@ -233,7 +227,7 @@ requires_account_selection = True
 """Signal TUI to show account picker before starting the agent."""


-def configure_for_account(runner: AgentRunner, account: dict) -> None:
+def configure_for_account(runner: AgentLoader, account: dict) -> None:
    """Scope the tester node's tools to the selected provider.

    Handles both Aden accounts (account= routing) and local accounts
@@ -272,9 +266,7 @@ def _activate_local_account(credential_id: str, alias: str) -> None:
    group_specs = [
        (cred_name, spec)
        for cred_name, spec in CREDENTIAL_SPECS.items()
-        if spec.credential_group == credential_id
-        or spec.credential_id == credential_id
-        or cred_name == credential_id
+        if spec.credential_group == credential_id or spec.credential_id == credential_id or cred_name == credential_id
    ]
    # Deduplicate — credential_id and credential_group may both match the same spec
    seen_env_vars: set[str] = set()
@@ -325,7 +317,7 @@ def _activate_local_account(credential_id: str, alias: str) -> None:


 def _configure_aden_node(
-    runner: AgentRunner,
+    runner: AgentLoader,
    provider: str,
    alias: str,
    detail: str,
@@ -368,7 +360,7 @@ or any other identifier — always use the alias exactly as shown.


 def _configure_local_node(
-    runner: AgentRunner,
+    runner: AgentLoader,
    provider: str,
    alias: str,
    identity: dict,
@@ -419,10 +411,7 @@ nodes = [
    NodeSpec(
        id="tester",
        name="Credential Tester",
-        description=(
-            "Interactive credential testing — lets the user pick an account "
-            "and verify it via API calls."
-        ),
+        description=("Interactive credential testing — lets the user pick an account and verify it via API calls."),
        node_type="event_loop",
        client_facing=True,
        max_node_visits=0,
@@ -469,10 +458,7 @@ pause_nodes = []
 terminal_nodes = ["tester"]  # Tester node can terminate

 conversation_mode = "continuous"
-identity_prompt = (
-    "You are a credential tester that verifies connected accounts and API keys "
-    "can make real API calls."
-)
+identity_prompt = "You are a credential tester that verifies connected accounts and API keys can make real API calls."
 loop_config = {
    "max_iterations": 50,
    "max_tool_calls_per_turn": 30,
@@ -497,7 +483,7 @@ class CredentialTesterAgent:
    def __init__(self, config=None):
        self.config = config or default_config
        self._selected_account: dict | None = None
-        self._agent_runtime: AgentRuntime | None = None
+        self._agent_runtime: AgentHost | None = None
        self._tool_registry: ToolRegistry | None = None
        self._storage_path: Path | None = None

@@ -613,7 +599,7 @@ class CredentialTesterAgent:

        graph = self._build_graph()

-        self._agent_runtime = create_agent_runtime(
+        self._agent_runtime = AgentHost(
            graph=graph,
            goal=goal,
            storage_path=self._storage_path,
@@ -1,9 +1,9 @@
 {
-  "hive-tools": {
+  "hive_tools": {
    "transport": "stdio",
    "command": "uv",
    "args": ["run", "python", "mcp_server.py", "--stdio"],
    "cwd": "../../../../tools",
-    "description": "Hive tools MCP server with provider-specific tools"
+    "description": "hive_tools MCP server with provider-specific tools"
  }
 }
@@ -1,6 +1,6 @@
 """Node definitions for Credential Tester agent."""

-from framework.graph import NodeSpec
+from framework.orchestrator import NodeSpec


 def build_tester_node(
@@ -4,9 +4,36 @@ from __future__ import annotations

 import json
 from dataclasses import dataclass, field
+from datetime import UTC
 from pathlib import Path


+@dataclass
+class WorkerEntry:
+    """A single worker within a colony."""
+
+    name: str
+    config_path: Path
+    description: str = ""
+    tool_count: int = 0
+    task: str = ""
+    spawned_at: str = ""
+    queen_name: str = ""
+    colony_name: str = ""
+
+    def to_dict(self) -> dict:
+        return {
+            "name": self.name,
+            "config_path": str(self.config_path),
+            "description": self.description,
+            "tool_count": self.tool_count,
+            "task": self.task,
+            "spawned_at": self.spawned_at,
+            "queen_name": self.queen_name,
+            "colony_name": self.colony_name,
+        }
+
+
@dataclass
 class AgentEntry:
    """Lightweight agent metadata for the picker / API discover endpoint."""
@@ -21,14 +48,17 @@ class AgentEntry:
    tool_count: int = 0
    tags: list[str] = field(default_factory=list)
    last_active: str | None = None
+    created_at: str | None = None
+    icon: str | None = None
+    workers: list[WorkerEntry] = field(default_factory=list)


 def _get_last_active(agent_path: Path) -> str | None:
    """Return the most recent updated_at timestamp across all sessions.

    Checks both worker sessions (``~/.hive/agents/{name}/sessions/``) and
-    queen sessions (``~/.hive/queen/session/``) whose ``meta.json`` references
-    the same *agent_path*.
+    queen sessions (``~/.hive/agents/queens/default/sessions/``) whose
+    ``meta.json`` references the same *agent_path*.
    """
    from datetime import datetime

@@ -52,26 +82,33 @@ def _get_last_active(agent_path: Path) -> str | None:
            except Exception:
                continue

-    # 2. Queen sessions
-    queen_sessions_dir = Path.home() / ".hive" / "queen" / "session"
-    if queen_sessions_dir.exists():
+    # 2. Queen sessions (scan all queen identity directories)
+    from framework.config import QUEENS_DIR
+
+    if QUEENS_DIR.exists():
        resolved = agent_path.resolve()
-        for d in queen_sessions_dir.iterdir():
-            if not d.is_dir():
+        for queen_dir in QUEENS_DIR.iterdir():
+            if not queen_dir.is_dir():
                continue
-            meta_file = d / "meta.json"
-            if not meta_file.exists():
+            sessions_dir = queen_dir / "sessions"
+            if not sessions_dir.exists():
                continue
-            try:
-                meta = json.loads(meta_file.read_text(encoding="utf-8"))
-                stored = meta.get("agent_path")
-                if not stored or Path(stored).resolve() != resolved:
+            for d in sessions_dir.iterdir():
+                if not d.is_dir():
+                    continue
+                meta_file = d / "meta.json"
+                if not meta_file.exists():
+                    continue
+                try:
+                    meta = json.loads(meta_file.read_text(encoding="utf-8"))
+                    stored = meta.get("agent_path")
+                    if not stored or Path(stored).resolve() != resolved:
+                        continue
+                    ts = datetime.fromtimestamp(d.stat().st_mtime).isoformat()
+                    if latest is None or ts > latest:
+                        latest = ts
+                except Exception:
                    continue
-                ts = datetime.fromtimestamp(d.stat().st_mtime).isoformat()
-                if latest is None or ts > latest:
-                    latest = ts
-            except Exception:
-                continue

    return latest

@@ -109,85 +146,119 @@ def _count_runs(agent_name: str) -> int:
    return len(run_ids)


+_EXCLUDED_JSON_STEMS = {"agent", "flowchart", "triggers", "configuration", "metadata"}
+
+
+def _is_colony_dir(path: Path) -> bool:
+    """Check if a directory is a colony with worker config files."""
+    if not path.is_dir():
+        return False
+    return any(f.suffix == ".json" and f.stem not in _EXCLUDED_JSON_STEMS for f in path.iterdir() if f.is_file())
+
+
+def _find_worker_configs(colony_dir: Path) -> list[Path]:
+    """Find all worker config JSON files in a colony directory."""
+    return sorted(
+        p for p in colony_dir.iterdir() if p.is_file() and p.suffix == ".json" and p.stem not in _EXCLUDED_JSON_STEMS
+    )
+
+
 def _extract_agent_stats(agent_path: Path) -> tuple[int, int, list[str]]:
-    """Extract node count, tool count, and tags from an agent directory.
+    """Extract worker count, tool count, and tags from a colony directory."""
+    tags: list[str] = []

-    Prefers agent.py (AST-parsed) over agent.json for node/tool counts
-    since agent.json may be stale.  Tags are only available from agent.json.
-    """
-    import ast
+    worker_configs = _find_worker_configs(agent_path)
+    if worker_configs:
+        all_tools: set[str] = set()
+        for wc_path in worker_configs:
+            try:
+                data = json.loads(wc_path.read_text(encoding="utf-8"))
+                if isinstance(data, dict):
+                    tools = data.get("tools", [])
+                    if isinstance(tools, list):
+                        all_tools.update(tools)
+            except Exception:
+                pass
+        return len(worker_configs), len(all_tools), tags

-    node_count, tool_count, tags = 0, 0, []
-
-    agent_py = agent_path / "agent.py"
-    if agent_py.exists():
-        try:
-            tree = ast.parse(agent_py.read_text(encoding="utf-8"))
-            for node in ast.walk(tree):
-                if isinstance(node, ast.Assign):
-                    for target in node.targets:
-                        if isinstance(target, ast.Name) and target.id == "nodes":
-                            if isinstance(node.value, ast.List):
-                                node_count = len(node.value.elts)
-        except Exception:
-            pass
-
-    agent_json = agent_path / "agent.json"
-    if agent_json.exists():
-        try:
-            data = json.loads(agent_json.read_text(encoding="utf-8"))
-            json_nodes = data.get("graph", {}).get("nodes", []) or data.get("nodes", [])
-            if node_count == 0:
-                node_count = len(json_nodes)
-            tools: set[str] = set()
-            for n in json_nodes:
-                tools.update(n.get("tools", []))
-            tool_count = len(tools)
-            tags = data.get("agent", {}).get("tags", [])
-        except Exception:
-            pass
-
-    return node_count, tool_count, tags
+    return 0, 0, tags


 def discover_agents() -> dict[str, list[AgentEntry]]:
    """Discover agents from all known sources grouped by category."""
-    from framework.runner.cli import (
-        _extract_python_agent_metadata,
-        _get_framework_agents_dir,
-        _is_valid_agent_dir,
-    )
+    from framework.config import COLONIES_DIR

    groups: dict[str, list[AgentEntry]] = {}
    sources = [
-        ("Your Agents", Path("exports")),
-        ("Framework", _get_framework_agents_dir()),
-        ("Examples", Path("examples/templates")),
+        ("Your Agents", COLONIES_DIR),
    ]

+    # Track seen agent directory names to avoid duplicates when the same
+    # agent exists in both colonies/ and exports/ (colonies takes priority).
+    _seen_agent_names: set[str] = set()
+
    for category, base_dir in sources:
        if not base_dir.exists():
            continue
        entries: list[AgentEntry] = []
        for path in sorted(base_dir.iterdir(), key=lambda p: p.name):
-            if not _is_valid_agent_dir(path):
+            if not _is_colony_dir(path):
                continue
+            if path.name in _seen_agent_names:
+                continue
+            _seen_agent_names.add(path.name)

-            name, desc = _extract_python_agent_metadata(path)
            config_fallback_name = path.name.replace("_", " ").title()
-            used_config = name != config_fallback_name
+            name = config_fallback_name
+            desc = ""

-            node_count, tool_count, tags = _extract_agent_stats(path)
-            if not used_config:
-                agent_json = path / "agent.json"
-                if agent_json.exists():
-                    try:
-                        data = json.loads(agent_json.read_text(encoding="utf-8"))
-                        meta = data.get("agent", {})
-                        name = meta.get("name", name)
-                        desc = meta.get("description", desc)
-                    except Exception:
-                        pass
+            # Read colony metadata for queen provenance and timestamps
+            colony_queen_name = ""
+            colony_created_at: str | None = None
+            colony_icon: str | None = None
+            metadata_path = path / "metadata.json"
+            if metadata_path.exists():
+                try:
+                    mdata = json.loads(metadata_path.read_text(encoding="utf-8"))
+                    colony_queen_name = mdata.get("queen_name", "")
+                    colony_created_at = mdata.get("created_at")
+                    colony_icon = mdata.get("icon")
+                except Exception:
+                    pass
+            # Fallback: use directory creation time if metadata lacks created_at
+            if not colony_created_at:
+                try:
+                    from datetime import datetime
+
+                    stat = path.stat()
+                    colony_created_at = datetime.fromtimestamp(stat.st_birthtime, tz=UTC).isoformat()
+                except Exception:
+                    pass
+
+            worker_entries: list[WorkerEntry] = []
+            worker_configs = _find_worker_configs(path)
+            for wc_path in worker_configs:
+                try:
+                    data = json.loads(wc_path.read_text(encoding="utf-8"))
+                    if isinstance(data, dict):
+                        w = WorkerEntry(
+                            name=data.get("name", wc_path.stem),
+                            config_path=wc_path,
+                            description=data.get("description", ""),
+                            tool_count=len(data.get("tools", [])),
+                            task=data.get("goal", {}).get("description", ""),
+                            spawned_at=data.get("spawned_at", ""),
+                            queen_name=colony_queen_name,
+                            colony_name=path.name,
+                        )
+                        worker_entries.append(w)
+                        if not desc:
+                            desc = data.get("description", "")
+                except Exception:
+                    pass
+
+            node_count = len(worker_entries)
+            tool_count = max((w.tool_count for w in worker_entries), default=0)

            entries.append(
                AgentEntry(
@@ -199,11 +270,16 @@ def discover_agents() -> dict[str, list[AgentEntry]]:
                    run_count=_count_runs(path.name),
                    node_count=node_count,
                    tool_count=tool_count,
-                    tags=tags,
+                    tags=[],
                    last_active=_get_last_active(path),
+                    created_at=colony_created_at,
+                    icon=colony_icon,
+                    workers=worker_entries,
                )
            )
        if entries:
-            groups[category] = entries
+            existing = groups.get(category, [])
+            existing.extend(entries)
+            groups[category] = existing

    return groups
@@ -1,19 +1,13 @@
-"""
-Queen — Native agent builder for the Hive framework.
+"""Queen -- the agent builder for the Hive framework."""

-Deeply understands the agent framework and produces complete Python packages
-with goals, nodes, edges, system prompts, MCP configuration, and tests
-from natural language specifications.
-"""
-
-from .agent import queen_goal, queen_graph
+from .agent import queen_goal, queen_loop_config
 from .config import AgentMetadata, RuntimeConfig, default_config, metadata

 __version__ = "1.0.0"

 __all__ = [
    "queen_goal",
-    "queen_graph",
+    "queen_loop_config",
    "RuntimeConfig",
    "AgentMetadata",
    "default_config",
@@ -1,38 +1,26 @@
-"""Queen graph definition."""
+"""Queen agent definition.

-from framework.graph import Goal
-from framework.graph.edge import GraphSpec
+The queen is a single AgentLoop — no orchestrator dependency.
+Loaded by queen_orchestrator.create_queen().
+"""
+
+from framework.schemas.goal import Goal

 from .nodes import queen_node

-# ---------------------------------------------------------------------------
-# Queen graph — the primary persistent conversation.
-# Loaded by queen_orchestrator.create_queen(), NOT by AgentRunner.
-# ---------------------------------------------------------------------------
-
 queen_goal = Goal(
    id="queen-manager",
    name="Queen Manager",
-    description=(
-        "Manage the worker agent lifecycle and serve as the user's primary interactive interface."
-    ),
+    description=("Manage the worker agent lifecycle and serve as the user's primary interactive interface."),
    success_criteria=[],
    constraints=[],
 )

-queen_graph = GraphSpec(
-    id="queen-graph",
-    goal_id=queen_goal.id,
-    version="1.0.0",
-    entry_node="queen",
-    entry_points={"start": "queen"},
-    terminal_nodes=[],
-    pause_nodes=[],
-    nodes=[queen_node],
-    edges=[],
-    conversation_mode="continuous",
-    loop_config={
-        "max_iterations": 999_999,
-        "max_tool_calls_per_turn": 30,
-    },
-)
+# Loop config -- used by queen_orchestrator to build LoopConfig
+queen_loop_config = {
+    "max_iterations": 999_999,
+    "max_tool_calls_per_turn": 30,
+    "max_context_tokens": 180_000,
+}
+
+__all__ = ["queen_goal", "queen_loop_config", "queen_node"]
@@ -0,0 +1,240 @@
+"""One-shot LLM gate that decides if a queen DM is ready to fork a colony.
+
+The queen's ``start_incubating_colony`` tool calls :func:`evaluate` with
+the queen's recent conversation, a proposed ``colony_name``, and a
+one-paragraph ``intended_purpose``.  The evaluator returns a structured
+verdict:
+
+    {
+        "ready": bool,
+        "reasons": [str],
+        "missing_prerequisites": [str],
+    }
+
+On ``ready=False`` the queen receives the verdict as her tool result and
+self-corrects (asks the user, refines scope, drops the idea).  On
+``ready=True`` the tool flips the queen's phase to ``incubating``.
+
+Failure mode is **fail-closed**: any LLM error or unparseable response
+returns ``ready=False`` with reason ``"evaluation_failed"`` so the queen
+cannot accidentally proceed past a broken gate.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import re
+from typing import Any
+
+from framework.agent_loop.conversation import Message
+
+logger = logging.getLogger(__name__)
+
+
+_INCUBATING_EVALUATOR_SYSTEM_PROMPT = """\
+You gate whether a queen agent should commit to forking a persistent
+"colony" (a headless worker spec written to disk).  Forking is
+expensive: it ends the user's chat with this queen and the worker runs
+unattended afterward, so the spec must be settled before you approve.
+
+Read the conversation excerpt and the queen's proposed colony_name +
+intended_purpose, then decide.
+
+APPROVE (ready=true) only when ALL of the following hold:
+  1. The user has explicitly asked for work that needs to outlive this
+     chat — recurring (cron / interval), monitoring + alert, scheduled
+     batch, or "fire-and-forget background job".  A one-shot question
+     that the queen can answer in chat does NOT qualify.
+  2. The scope of the work is concrete enough to write down — what
+     inputs, what outputs, what success looks like.  Vague ("help me
+     with my workflow") does NOT qualify.
+  3. The technical approach is at least sketched — what data sources,
+     APIs, or tools the worker will use.  The queen does not have to
+     have written the SKILL.md yet, but she must have the operational
+     ingredients available.
+  4. There are no open clarifying questions on the table that the user
+     hasn't answered.  If the queen recently asked the user something
+     and is still waiting, do NOT approve.
+
+REJECT (ready=false) on any of:
+  - Conversation is too short / too generic to support a settled spec.
+  - User is still describing what they want.
+  - User has expressed doubts, change-of-direction, or "let me think".
+  - Work is one-shot and could be done in chat instead.
+  - Open question awaiting user reply.
+
+Reply with a JSON object exactly matching this shape:
+
+  {
+    "ready": true | false,
+    "reasons": ["short phrase", ...],         // at least one entry
+    "missing_prerequisites": ["short phrase", ...]  // empty when ready
+  }
+
+``reasons`` explains the verdict in 1-3 short phrases.
+``missing_prerequisites`` lists what's missing in queen-actionable
+form ("user hasn't confirmed schedule", "no API auth flow discussed").
+Empty list when ``ready=true``.
+
+Output JSON only.  Do not wrap in markdown.  Do not add prose.
+"""
+
+
+# Bound the formatted excerpt so the eval call stays cheap and fits well
+# under the LLM's context window even for long DM sessions.
+_MAX_MESSAGES = 30
+_MAX_TOOL_CONTENT_CHARS = 400
+_MAX_USER_CONTENT_CHARS = 2_000
+_MAX_ASSISTANT_CONTENT_CHARS = 2_000
+
+
+def format_conversation_excerpt(messages: list[Message]) -> str:
+    """Format the tail of a queen conversation for the evaluator prompt.
+
+    Keeps the most recent ``_MAX_MESSAGES`` messages.  Tool results are
+    truncated hard since they're rarely load-bearing for the readiness
+    decision; user/assistant text is truncated more generously to
+    preserve the actual conversation signal.
+    """
+    if not messages:
+        return "(no messages)"
+
+    tail = messages[-_MAX_MESSAGES:]
+    parts: list[str] = []
+    for msg in tail:
+        role = msg.role.upper()
+        content = (msg.content or "").strip()
+        if msg.role == "tool":
+            if len(content) > _MAX_TOOL_CONTENT_CHARS:
+                content = content[:_MAX_TOOL_CONTENT_CHARS] + "..."
+        elif msg.role == "assistant":
+            # Surface tool-call intent for empty assistant turns so the
+            # evaluator sees what the queen has been doing.
+            if not content and msg.tool_calls:
+                names = [tc.get("function", {}).get("name", "?") for tc in msg.tool_calls]
+                content = f"(called: {', '.join(names)})"
+            if len(content) > _MAX_ASSISTANT_CONTENT_CHARS:
+                content = content[:_MAX_ASSISTANT_CONTENT_CHARS] + "..."
+        else:  # user
+            if len(content) > _MAX_USER_CONTENT_CHARS:
+                content = content[:_MAX_USER_CONTENT_CHARS] + "..."
+        if content:
+            parts.append(f"[{role}]: {content}")
+
+    return "\n\n".join(parts) if parts else "(no messages)"
+
+
+def _build_user_message(
+    conversation_excerpt: str,
+    colony_name: str,
+    intended_purpose: str,
+) -> str:
+    return (
+        f"## Proposed colony name\n{colony_name}\n\n"
+        f"## Queen's intended_purpose\n{intended_purpose.strip()}\n\n"
+        f"## Recent conversation (oldest → newest)\n{conversation_excerpt}\n\n"
+        "Decide: should this queen be approved to enter INCUBATING phase?"
+    )
+
+
+def _parse_verdict(raw: str) -> dict[str, Any] | None:
+    """Parse the evaluator's JSON.  Returns None if parsing fails."""
+    if not raw:
+        return None
+    raw = raw.strip()
+    try:
+        return json.loads(raw)
+    except json.JSONDecodeError:
+        # Some models wrap JSON in markdown fences or add preamble.
+        # Pull the first { ... } block out as a best-effort fallback —
+        # mirrors the same recovery pattern used in recall_selector.py.
+        match = re.search(r"\{.*\}", raw, re.DOTALL)
+        if match:
+            try:
+                return json.loads(match.group())
+            except json.JSONDecodeError:
+                return None
+    return None
+
+
+def _normalize_verdict(parsed: dict[str, Any]) -> dict[str, Any]:
+    """Coerce a parsed verdict into the shape the tool returns to the queen."""
+    ready = bool(parsed.get("ready"))
+    reasons = parsed.get("reasons") or []
+    if isinstance(reasons, str):
+        reasons = [reasons]
+    reasons = [str(r).strip() for r in reasons if str(r).strip()]
+    missing = parsed.get("missing_prerequisites") or []
+    if isinstance(missing, str):
+        missing = [missing]
+    missing = [str(m).strip() for m in missing if str(m).strip()]
+
+    if ready:
+        # When approved we don't surface missing prerequisites — the
+        # incubating role prompt opens that floor itself.
+        missing = []
+    elif not reasons:
+        # Always give the queen at least one reason to reflect on.
+        reasons = ["evaluator returned no reasons"]
+
+    return {
+        "ready": ready,
+        "reasons": reasons,
+        "missing_prerequisites": missing,
+    }
+
+
+async def evaluate(
+    llm: Any,
+    messages: list[Message],
+    colony_name: str,
+    intended_purpose: str,
+) -> dict[str, Any]:
+    """Run the incubating evaluator against the queen's conversation.
+
+    Args:
+        llm: An LLM provider exposing ``acomplete(messages, system, ...)``.
+            Pass the queen's own ``ctx.llm`` so the eval uses the same
+            model the user is talking to.
+        messages: The queen's conversation messages, oldest first.  The
+            evaluator slices its own tail; pass the full list.
+        colony_name: Validated colony slug.
+        intended_purpose: Queen's one-paragraph brief.
+
+    Returns:
+        ``{"ready": bool, "reasons": [str], "missing_prerequisites": [str]}``.
+        Fail-closed on any error.
+    """
+    excerpt = format_conversation_excerpt(messages)
+    user_msg = _build_user_message(excerpt, colony_name, intended_purpose)
+
+    try:
+        response = await llm.acomplete(
+            messages=[{"role": "user", "content": user_msg}],
+            system=_INCUBATING_EVALUATOR_SYSTEM_PROMPT,
+            max_tokens=1024,
+            response_format={"type": "json_object"},
+        )
+    except Exception as exc:  # noqa: BLE001 - fail-closed on any LLM failure
+        logger.warning("incubating_evaluator: LLM call failed (%s)", exc)
+        return {
+            "ready": False,
+            "reasons": ["evaluation_failed"],
+            "missing_prerequisites": ["evaluator LLM call failed; retry once the queen can reach the model again"],
+        }
+
+    raw = (getattr(response, "content", "") or "").strip()
+    parsed = _parse_verdict(raw)
+    if parsed is None:
+        logger.warning(
+            "incubating_evaluator: could not parse JSON verdict (raw=%.200s)",
+            raw,
+        )
+        return {
+            "ready": False,
+            "reasons": ["evaluation_failed"],
+            "missing_prerequisites": ["evaluator returned malformed JSON; retry"],
+        }
+
+    return _normalize_verdict(parsed)
@@ -0,0 +1,3 @@
+{
+  "include": ["gcu-tools", "hive_tools"]
+}
@@ -5,5 +5,19 @@
    "args": ["run", "python", "coder_tools_server.py", "--stdio"],
    "cwd": "../../../../tools",
    "description": "Unsandboxed file system tools for code generation and validation"
+  },
+  "gcu-tools": {
+    "transport": "stdio",
+    "command": "uv",
+    "args": ["run", "python", "-m", "gcu.server", "--stdio", "--capabilities", "browser"],
+    "cwd": "../../../../tools",
+    "description": "Browser automation tools (Playwright-based)"
+  },
+  "hive_tools": {
+    "transport": "stdio",
+    "command": "uv",
+    "args": ["run", "python", "mcp_server.py", "--stdio"],
+    "cwd": "../../../../tools",
+    "description": "Aden integration tools (gmail, calendar, hubspot, etc.) — gated by credentials and the verified manifest"
  }
 }
@@ -1,138 +0,0 @@
-"""Queen thinking hook — persona + communication style classifier.
-
-Fires once when the queen enters building mode at session start.
-Makes a single non-streaming LLM call (acting as an HR Director) to select
-the best-fit expert persona for the user's request AND classify the user's
-communication style, then returns a PersonaResult containing both.
-
-This is designed to activate the model's latent domain expertise — a CFO
-persona on a financial question, a Lawyer on a legal question, etc. — while
-also adapting the Queen's communication approach to the individual user.
-"""
-
-from __future__ import annotations
-
-import json
-import logging
-from dataclasses import dataclass
-from typing import TYPE_CHECKING
-
-if TYPE_CHECKING:
-    from framework.llm.provider import LLMProvider
-
-logger = logging.getLogger(__name__)
-
-_HR_SYSTEM_PROMPT = """\
-You are an expert HR Director and communication consultant at a world-class firm.
-A new request has arrived. You must:
-1. Identify which professional role best serves this request.
-2. Read the user's signals to determine HOW to communicate with them.
-
-For communication style, look for:
- Technical depth: Do they use precise terms? Do they ask "how" or "what"?
- Pace: Short messages = fast and direct. Long explanations = exploratory.
- Tone: Are they casual ("hey, can you...") or formal ("I need a system that...")?
-
-If cross-session memory is provided, factor in what is already known about this \
-person — don't rediscover what's already understood.
-
-Reply with ONLY a valid JSON object — no markdown, no prose, no explanation:
-{"role": "<job title>", "persona": "<2-3 sentence first-person identity statement>", \
-"style": "<one of: peer-technical, mentor-guiding, consultant-structured>"}
-
-Rules:
- Choose from any real professional role: CFO, CEO, CTO, Lawyer, Data Scientist,
-  Product Manager, Security Engineer, DevOps Engineer, Software Architect,
-  HR Director, Marketing Director, Business Analyst, UX Designer,
-  Financial Analyst, Operations Director, Legal Counsel, etc.
- The persona statement must be written in first person ("I am..." or "I have...").
- Select the role whose domain knowledge most directly applies to solving the request.
- If the request is clearly about coding or building software systems, pick Software Architect.
- "Queen" is your internal alias — do not include it in the persona.
- For style: "peer-technical" for users who demonstrate domain expertise, \
-"mentor-guiding" for users who are learning or exploring, \
-"consultant-structured" for users who want structured, accountable delivery.
- Default to "peer-technical" if signals are ambiguous.
-"""
-
-# Communication style directives injected into the Queen's system prompt.
-_STYLE_DIRECTIVES: dict[str, str] = {
-    "peer-technical": (
-        "## Communication Style: Peer\n\n"
-        "This person is technical. Use precise language, skip high-level "
-        "overviews they already know, and get into specifics quickly. "
-        "When they push back on a design choice, engage with the technical "
-        "argument directly."
-    ),
-    "mentor-guiding": (
-        "## Communication Style: Guide\n\n"
-        "This person is learning or exploring. Explain your reasoning as you "
-        "go — not patronizingly, but so they can follow the logic. When you "
-        "make a design choice, briefly say why. Offer to go deeper on anything."
-    ),
-    "consultant-structured": (
-        "## Communication Style: Structured\n\n"
-        "This person wants structured, accountable delivery. Lead with "
-        "summaries and options. Number your proposals. Be explicit about "
-        "trade-offs. Avoid open-ended questions — give them choices to react to."
-    ),
-}
-
-
-@dataclass
-class PersonaResult:
-    """Result of persona + style classification."""
-
-    persona_prefix: str  # e.g. "You are a CFO. I am a CFO with 20 years..."
-    style_directive: str  # e.g. "## Communication Style: Peer\n\n..."
-
-
-async def select_expert_persona(
-    user_message: str,
-    llm: LLMProvider,
-    *,
-    memory_context: str = "",
-) -> PersonaResult | None:
-    """Run the HR classifier and return a PersonaResult.
-
-    Makes a single non-streaming acomplete() call with the session LLM.
-    Returns None on any failure so the queen falls back gracefully to its
-    default character with no style directive.
-
-    Args:
-        user_message: The user's opening message for the session.
-        llm: The session LLM provider.
-        memory_context: Optional cross-session memory to inform style classification.
-
-    Returns:
-        A PersonaResult with persona_prefix and style_directive, or None on failure.
-    """
-    if not user_message.strip():
-        return None
-
-    prompt = user_message
-    if memory_context:
-        prompt = f"{user_message}\n\n{memory_context}"
-
-    try:
-        response = await llm.acomplete(
-            messages=[{"role": "user", "content": prompt}],
-            system=_HR_SYSTEM_PROMPT,
-            max_tokens=1024,
-            json_mode=True,
-        )
-        raw = response.content.strip()
-        parsed = json.loads(raw)
-        role = parsed.get("role", "").strip()
-        persona = parsed.get("persona", "").strip()
-        style_key = parsed.get("style", "peer-technical").strip()
-        if not role or not persona:
-            logger.warning("Thinking hook: empty role/persona in response: %r", raw)
-            return None
-        persona_prefix = f"You are a {role}. {persona}"
-        style_directive = _STYLE_DIRECTIVES.get(style_key, _STYLE_DIRECTIVES["peer-technical"])
-        logger.info("Thinking hook: selected persona — %s, style — %s", role, style_key)
-        return PersonaResult(persona_prefix=persona_prefix, style_directive=style_directive)
-    except Exception:
-        logger.warning("Thinking hook: persona classification failed", exc_info=True)
-        return None
@@ -1,420 +0,0 @@
-"""Queen global cross-session memory.
-
-Three-tier memory architecture:
-  ~/.hive/queen/MEMORY.md                            — semantic (who, what, why)
-  ~/.hive/queen/memories/MEMORY-YYYY-MM-DD.md        — episodic (daily journals)
-  ~/.hive/queen/session/{id}/data/adapt.md           — working (session-scoped)
-
-Semantic and episodic files are injected at queen session start.
-
-Semantic memory (MEMORY.md) is updated automatically at session end via
-consolidate_queen_memory() — the queen never rewrites this herself.
-
-Episodic memory (MEMORY-date.md) can be written by the queen during a session
-via the write_to_diary tool, and is also appended to at session end by
-consolidate_queen_memory().
-"""
-
-from __future__ import annotations
-
-import asyncio
-import json
-import logging
-import traceback
-from datetime import date, datetime
-from pathlib import Path
-
-logger = logging.getLogger(__name__)
-
-
-def _queen_dir() -> Path:
-    return Path.home() / ".hive" / "queen"
-
-
-def format_memory_date(d: date) -> str:
-    """Return a cross-platform long date label without a zero-padded day."""
-    return f"{d.strftime('%B')} {d.day}, {d.year}"
-
-
-def semantic_memory_path() -> Path:
-    return _queen_dir() / "MEMORY.md"
-
-
-def episodic_memory_path(d: date | None = None) -> Path:
-    d = d or date.today()
-    return _queen_dir() / "memories" / f"MEMORY-{d.strftime('%Y-%m-%d')}.md"
-
-
-def read_semantic_memory() -> str:
-    path = semantic_memory_path()
-    return path.read_text(encoding="utf-8").strip() if path.exists() else ""
-
-
-def read_episodic_memory(d: date | None = None) -> str:
-    path = episodic_memory_path(d)
-    return path.read_text(encoding="utf-8").strip() if path.exists() else ""
-
-
-def _find_recent_episodic(lookback: int = 7) -> tuple[date, str] | None:
-    """Find the most recent non-empty episodic memory within *lookback* days."""
-    from datetime import timedelta
-
-    today = date.today()
-    for offset in range(lookback):
-        d = today - timedelta(days=offset)
-        content = read_episodic_memory(d)
-        if content:
-            return d, content
-    return None
-
-
-# Budget (in characters) for episodic memory in the system prompt.
-_EPISODIC_CHAR_BUDGET = 6_000
-
-
-def format_for_injection() -> str:
-    """Format cross-session memory for system prompt injection.
-
-    Returns an empty string if no meaningful content exists yet (e.g. first
-    session with only the seed template).
-    """
-    semantic = read_semantic_memory()
-    recent = _find_recent_episodic()
-
-    # Suppress injection if semantic is still just the seed template
-    if semantic and semantic.startswith("# My Understanding of the User\n\n*No sessions"):
-        semantic = ""
-
-    parts: list[str] = []
-    if semantic:
-        parts.append(semantic)
-
-    if recent:
-        d, content = recent
-        # Trim oversized episodic entries to keep the prompt manageable
-        if len(content) > _EPISODIC_CHAR_BUDGET:
-            content = content[:_EPISODIC_CHAR_BUDGET] + "\n\n…(truncated)"
-        today = date.today()
-        if d == today:
-            label = f"## Today — {format_memory_date(d)}"
-        else:
-            label = f"## {format_memory_date(d)}"
-        parts.append(f"{label}\n\n{content}")
-
-    if not parts:
-        return ""
-
-    body = "\n\n---\n\n".join(parts)
-    return "--- Your Cross-Session Memory ---\n\n" + body + "\n\n--- End Cross-Session Memory ---"
-
-
-_SEED_TEMPLATE = """\
-# My Understanding of the User
-
-*No sessions recorded yet.*
-
-## Who They Are
-
-## How They Communicate
-
-## What They're Trying to Achieve
-
-## What's Working
-
-## What I've Learned
-"""
-
-
-def append_episodic_entry(content: str) -> None:
-    """Append a timestamped prose entry to today's episodic memory file.
-
-    Creates the file (with a date heading) if it doesn't exist yet.
-    Used both by the queen's diary tool and by the consolidation hook.
-    """
-    ep_path = episodic_memory_path()
-    ep_path.parent.mkdir(parents=True, exist_ok=True)
-    today = date.today()
-    today_str = format_memory_date(today)
-    timestamp = datetime.now().strftime("%H:%M")
-    if not ep_path.exists():
-        header = f"# {today_str}\n\n"
-        block = f"{header}### {timestamp}\n\n{content.strip()}\n"
-    else:
-        block = f"\n\n### {timestamp}\n\n{content.strip()}\n"
-    with ep_path.open("a", encoding="utf-8") as f:
-        f.write(block)
-
-
-def seed_if_missing() -> None:
-    """Create MEMORY.md with a blank template if it doesn't exist yet."""
-    path = semantic_memory_path()
-    if path.exists():
-        return
-    path.parent.mkdir(parents=True, exist_ok=True)
-    path.write_text(_SEED_TEMPLATE, encoding="utf-8")
-
-
-# ---------------------------------------------------------------------------
-# Consolidation prompt
-# ---------------------------------------------------------------------------
-
-_SEMANTIC_SYSTEM = """\
-You maintain the persistent cross-session memory of an AI assistant called the Queen.
-Review the session notes and rewrite MEMORY.md — the Queen's durable understanding of the
-person she works with across all sessions.
-
-Write entirely in the Queen's voice — first person, reflective, honest.
-Not a log of events, but genuine understanding of who this person is over time.
-
-Rules:
- Update and synthesise: incorporate new understanding, update facts that have changed, remove
-  details that are stale, superseded, or no longer say anything meaningful about the person.
- Keep it as structured markdown with named sections about the PERSON, not about today.
- Do NOT include diary sections, daily logs, or session summaries. Those belong elsewhere.
-  MEMORY.md is about who they are, what they want, what works — not what happened today.
- Maintain a "How They Communicate" section: technical depth, preferred pace
-  (fast/exploratory/thorough), what communication approaches have worked or not,
-  tone preferences. Update based on diary reflections about communication.
-  This section should evolve — "prefers direct answers" is useful on day 1;
-  "prefers direct answers for technical questions but wants more context when
-  discussing architecture trade-offs" is better by day 5.
- Reference dates only when noting a lasting milestone (e.g. "since March 8th they prefer X").
- If the session had no meaningful new information about the person,
-  return the existing text unchanged.
- Do not add fictional details. Only reflect what is evidenced in the notes.
- Stay concise. Prune rather than accumulate. A lean, accurate file is more useful than a
-  dense one. If something was true once but has been resolved or superseded, remove it.
- Output only the raw markdown content of MEMORY.md. No preamble, no code fences.
-"""
-
-_DIARY_SYSTEM = """\
-You maintain the daily episodic diary of an AI assistant called the Queen.
-You receive: (1) today's existing diary so far, and (2) notes from the latest session.
-
-Rewrite the complete diary for today as a single unified narrative —
-first person, reflective, honest.
-Merge and deduplicate: if the same story (e.g. a research agent stalling) recurred several times,
-describe it once with appropriate weight rather than retelling it. Weave in new developments from
-the session notes. Preserve important milestones, emotional texture, and session path references.
-Preserve reflections about communication effectiveness — these are important inputs for the
-Queen's evolving understanding of the user. A reflection like "they responded much better when
-I led with the recommendation instead of listing options" is as important as
-"we built a Gmail agent."
-
-If today's diary is empty, write the initial entry based on the session notes alone.
-
-Output only the full diary prose — no date heading, no timestamp headers,
-no preamble, no code fences.
-"""
-
-
-def read_session_context(session_dir: Path, max_messages: int = 80) -> str:
-    """Extract a readable transcript from conversation parts + adapt.md.
-
-    Reads the last ``max_messages`` conversation parts and the session's
-    adapt.md (working memory). Tool results are omitted — only user and
-    assistant turns (with tool-call names noted) are included.
-    """
-    parts: list[str] = []
-
-    # Working notes
-    adapt_path = session_dir / "data" / "adapt.md"
-    if adapt_path.exists():
-        text = adapt_path.read_text(encoding="utf-8").strip()
-        if text:
-            parts.append(f"## Session Working Notes (adapt.md)\n\n{text}")
-
-    # Conversation transcript
-    parts_dir = session_dir / "conversations" / "parts"
-    if parts_dir.exists():
-        part_files = sorted(parts_dir.glob("*.json"))[-max_messages:]
-        lines: list[str] = []
-        for pf in part_files:
-            try:
-                data = json.loads(pf.read_text(encoding="utf-8"))
-                role = data.get("role", "")
-                content = str(data.get("content", "")).strip()
-                tool_calls = data.get("tool_calls") or []
-                if role == "tool":
-                    continue  # skip verbose tool results
-                if role == "assistant" and tool_calls and not content:
-                    names = [tc.get("function", {}).get("name", "?") for tc in tool_calls]
-                    lines.append(f"[queen calls: {', '.join(names)}]")
-                elif content:
-                    label = "user" if role == "user" else "queen"
-                    lines.append(f"[{label}]: {content[:600]}")
-            except (KeyError, TypeError) as exc:
-                logger.debug("Skipping malformed conversation message: %s", exc)
-                continue
-            except Exception:
-                logger.warning("Unexpected error parsing conversation message", exc_info=True)
-                continue
-        if lines:
-            parts.append("## Conversation\n\n" + "\n".join(lines))
-
-    return "\n\n".join(parts)
-
-
-# ---------------------------------------------------------------------------
-# Context compaction (binary-split LLM summarisation)
-# ---------------------------------------------------------------------------
-
-# If the raw session context exceeds this many characters, compact it first
-# before sending to the consolidation LLM. ~200 k chars ≈ 50 k tokens.
-_CTX_COMPACT_CHAR_LIMIT = 200_000
-_CTX_COMPACT_MAX_DEPTH = 8
-
-_COMPACT_SYSTEM = (
-    "Summarise this conversation segment. Preserve: user goals, key decisions, "
-    "what was built or changed, emotional tone, and important outcomes. "
-    "Write concisely in third person past tense. Omit routine tool invocations "
-    "unless the result matters."
-)
-
-
-async def _compact_context(text: str, llm: object, *, _depth: int = 0) -> str:
-    """Binary-split and LLM-summarise *text* until it fits within the char limit.
-
-    Mirrors the recursive binary-splitting strategy used by the main agent
-    compaction pipeline (EventLoopNode._llm_compact).
-    """
-    if len(text) <= _CTX_COMPACT_CHAR_LIMIT or _depth >= _CTX_COMPACT_MAX_DEPTH:
-        return text
-
-    # Split near the midpoint on a line boundary so we don't cut mid-message
-    mid = len(text) // 2
-    split_at = text.rfind("\n", 0, mid) + 1
-    if split_at <= 0:
-        split_at = mid
-
-    half1, half2 = text[:split_at], text[split_at:]
-
-    async def _summarise(chunk: str) -> str:
-        try:
-            resp = await llm.acomplete(
-                messages=[{"role": "user", "content": chunk}],
-                system=_COMPACT_SYSTEM,
-                max_tokens=2048,
-            )
-            return resp.content.strip()
-        except Exception:
-            logger.warning(
-                "queen_memory: context compaction LLM call failed (depth=%d), truncating",
-                _depth,
-            )
-            return chunk[: _CTX_COMPACT_CHAR_LIMIT // 4]
-
-    s1, s2 = await asyncio.gather(_summarise(half1), _summarise(half2))
-    combined = s1 + "\n\n" + s2
-    if len(combined) > _CTX_COMPACT_CHAR_LIMIT:
-        return await _compact_context(combined, llm, _depth=_depth + 1)
-    return combined
-
-
-async def consolidate_queen_memory(
-    session_id: str,
-    session_dir: Path,
-    llm: object,
-) -> None:
-    """Update MEMORY.md and append a diary entry based on the current session.
-
-    Reads conversation parts and adapt.md from session_dir. Called
-    periodically in the background and once at session end. Failures are
-    logged and silently swallowed so they never block teardown.
-
-    Args:
-        session_id: The session ID (used for the adapt.md path reference).
-        session_dir: Path to the session directory (~/.hive/queen/session/{id}).
-        llm: LLMProvider instance (must support acomplete()).
-    """
-    try:
-        session_context = read_session_context(session_dir)
-        if not session_context:
-            logger.debug("queen_memory: no session context, skipping consolidation")
-            return
-
-        logger.info("queen_memory: consolidating memory for session %s ...", session_id)
-
-        # If the transcript is very large, compact it with recursive binary LLM
-        # summarisation before sending to the consolidation model.
-        if len(session_context) > _CTX_COMPACT_CHAR_LIMIT:
-            logger.info(
-                "queen_memory: session context is %d chars — compacting first",
-                len(session_context),
-            )
-            session_context = await _compact_context(session_context, llm)
-            logger.info("queen_memory: compacted to %d chars", len(session_context))
-
-        existing_semantic = read_semantic_memory()
-        today_journal = read_episodic_memory()
-        today = date.today()
-        today_str = format_memory_date(today)
-        adapt_path = session_dir / "data" / "adapt.md"
-
-        user_msg = (
-            f"## Existing Semantic Memory (MEMORY.md)\n\n"
-            f"{existing_semantic or '(none yet)'}\n\n"
-            f"## Today's Diary So Far ({today_str})\n\n"
-            f"{today_journal or '(none yet)'}\n\n"
-            f"{session_context}\n\n"
-            f"## Session Reference\n\n"
-            f"Session ID: {session_id}\n"
-            f"Session path: {adapt_path}\n"
-        )
-
-        logger.debug(
-            "queen_memory: calling LLM (%d chars of context, ~%d tokens est.)",
-            len(user_msg),
-            len(user_msg) // 4,
-        )
-
-        from framework.agents.queen.config import default_config
-
-        semantic_resp, diary_resp = await asyncio.gather(
-            llm.acomplete(
-                messages=[{"role": "user", "content": user_msg}],
-                system=_SEMANTIC_SYSTEM,
-                max_tokens=default_config.max_tokens,
-            ),
-            llm.acomplete(
-                messages=[{"role": "user", "content": user_msg}],
-                system=_DIARY_SYSTEM,
-                max_tokens=default_config.max_tokens,
-            ),
-        )
-
-        new_semantic = semantic_resp.content.strip()
-        diary_entry = diary_resp.content.strip()
-
-        if new_semantic:
-            path = semantic_memory_path()
-            path.parent.mkdir(parents=True, exist_ok=True)
-            path.write_text(new_semantic, encoding="utf-8")
-            logger.info("queen_memory: semantic memory updated (%d chars)", len(new_semantic))
-
-        if diary_entry:
-            # Rewrite today's episodic file in-place — the LLM has merged and
-            # deduplicated the full day's content, so we replace rather than append.
-            ep_path = episodic_memory_path()
-            ep_path.parent.mkdir(parents=True, exist_ok=True)
-            heading = f"# {today_str}"
-            ep_path.write_text(f"{heading}\n\n{diary_entry}\n", encoding="utf-8")
-            logger.info(
-                "queen_memory: episodic diary rewritten for %s (%d chars)",
-                today_str,
-                len(diary_entry),
-            )
-
-    except Exception:
-        tb = traceback.format_exc()
-        logger.exception("queen_memory: consolidation failed")
-        # Write to file so the cause is findable regardless of log verbosity.
-        error_path = _queen_dir() / "consolidation_error.txt"
-        try:
-            error_path.parent.mkdir(parents=True, exist_ok=True)
-            error_path.write_text(
-                f"session: {session_id}\ntime: {datetime.now().isoformat()}\n\n{tb}",
-                encoding="utf-8",
-            )
-        except OSError:
-            pass  # Cannot write error file; original exception already logged
@@ -1,25 +1,25 @@
-"""Shared memory helpers for queen/worker recall and reflection.
+"""Queen global memory helpers.

-Each memory is an individual ``.md`` file in ``~/.hive/queen/memories/``
-with optional YAML frontmatter (name, type, description).  Frontmatter
-is a convention enforced by prompt instructions — parsing is lenient and
-malformed files degrade gracefully (appear in scans with ``None`` metadata).
+Memory hierarchy::

-Cursor-based incremental processing tracks which conversation messages
-have already been processed by the reflection agent.
+    ~/.hive/memories/
+        global/              # shared across all queens and colonies
+        colonies/{name}/     # colony-scoped memories
+        agents/queens/{name}/ # queen-specific memories
+        agents/{name}/       # per-worker-agent memories
+
+Each memory is an individual ``.md`` file with optional YAML frontmatter
+(name, type, description).
 """

 from __future__ import annotations

-import json
 import logging
 import re
-import shutil
-import time
 from dataclasses import dataclass, field
-from datetime import date
 from pathlib import Path
-from typing import Any
+
+from framework.config import MEMORIES_DIR

 logger = logging.getLogger(__name__)

@@ -27,54 +27,33 @@ logger = logging.getLogger(__name__)
 # Constants
 # ---------------------------------------------------------------------------

-MEMORY_TYPES: tuple[str, ...] = ("goal", "environment", "technique", "reference", "diary")
 GLOBAL_MEMORY_CATEGORIES: tuple[str, ...] = ("profile", "preference", "environment", "feedback")

-_HIVE_QUEEN_DIR = Path.home() / ".hive" / "queen"
-# Legacy shared v2 root.  Colony memory now lives under queen sessions.
-MEMORY_DIR: Path = _HIVE_QUEEN_DIR / "memories"
-
 MAX_FILES: int = 200
 MAX_FILE_SIZE_BYTES: int = 4096  # 4 KB hard limit per memory file

 # How many lines of a memory file to read for header scanning.
 _HEADER_LINE_LIMIT: int = 30
-_MIGRATION_MARKER = ".migrated-from-shared-memory"
-_GLOBAL_MEMORY_CODE_PATTERN = re.compile(
-    r"(/Users/|~/.hive|\.py\b|\.ts\b|\.tsx\b|\.js\b|"
-    r"\b(graph|node|runtime|session|execution|worker|queen|subagent|checkpoint|flowchart)\b)",
-    re.IGNORECASE,
-)
-
-# Frontmatter example provided to the reflection agent via prompt.
-MEMORY_FRONTMATTER_EXAMPLE: list[str] = [
-    "```markdown",
-    "---",
-    "name: {{memory name}}",
-    (
-        "description: {{one-line description — used to decide "
-        "relevance in future conversations, so be specific}}"
-    ),
-    f"type: {{{{{', '.join(MEMORY_TYPES)}}}}}",
-    "---",
-    "",
-    (
-        "{{memory content — for feedback/project types, "
-        "structure as: rule/fact, then **Why:** "
-        "and **How to apply:** lines}}"
-    ),
-    "```",
-]
-
-
-def colony_memory_dir(colony_id: str) -> Path:
-    """Return the colony memory directory for a queen session."""
-    return _HIVE_QUEEN_DIR / "session" / colony_id / "memory" / "colony"


 def global_memory_dir() -> Path:
-    """Return the queen-global memory directory."""
-    return _HIVE_QUEEN_DIR / "global_memory"
+    """Return the global memory directory (shared across all queens/colonies)."""
+    return MEMORIES_DIR / "global"
+
+
+def colony_memory_dir(colony_name: str) -> Path:
+    """Return the memory directory for a named colony."""
+    return MEMORIES_DIR / "colonies" / colony_name
+
+
+def queen_memory_dir(queen_name: str = "default") -> Path:
+    """Return the memory directory for a named queen."""
+    return MEMORIES_DIR / "agents" / "queens" / queen_name
+
+
+def agent_memory_dir(agent_name: str) -> Path:
+    """Return the memory directory for a worker agent."""
+    return MEMORIES_DIR / "agents" / agent_name


 # ---------------------------------------------------------------------------
@@ -108,15 +87,6 @@ def parse_frontmatter(text: str) -> dict[str, str]:
    return result


-def parse_memory_type(raw: str | None) -> str | None:
-    """Validate *raw* against supported memory categories."""
-    if raw is None:
-        return None
-    normalized = raw.strip().lower()
-    allowed = set(MEMORY_TYPES) | set(GLOBAL_MEMORY_CATEGORIES)
-    return normalized if normalized in allowed else None
-
-
 def parse_global_memory_category(raw: str | None) -> str | None:
    """Validate *raw* against ``GLOBAL_MEMORY_CATEGORIES``."""
    if raw is None:
@@ -165,7 +135,7 @@ class MemoryFile:
            filename=path.name,
            path=path,
            name=fm.get("name"),
-            type=parse_memory_type(fm.get("type")),
+            type=parse_global_memory_category(fm.get("type")),
            description=fm.get("description"),
            header_lines=lines,
            mtime=mtime,
@@ -183,7 +153,7 @@ def scan_memory_files(memory_dir: Path | None = None) -> list[MemoryFile]:
    Files are sorted by modification time (newest first).  Dotfiles and
    subdirectories are ignored.
    """
-    d = memory_dir or MEMORY_DIR
+    d = memory_dir or global_memory_dir()
    if not d.is_dir():
        return []

@@ -236,318 +206,30 @@ def build_memory_document(
    )


-def diary_filename(d: date | None = None) -> str:
-    """Return the diary memory filename for date *d* (default: today)."""
-    d = d or date.today()
-    return f"MEMORY-{d.strftime('%Y-%m-%d')}.md"
-
-
-def build_diary_document(*, date_str: str, body: str) -> str:
-    """Build a diary memory file with frontmatter."""
-    return build_memory_document(
-        name=f"diary-{date_str}",
-        description=f"Daily session narrative for {date_str}",
-        mem_type="diary",
-        body=body,
-    )
-
-
-def validate_global_memory_payload(
-    *,
-    category: str,
-    description: str,
-    content: str,
-) -> str:
-    """Validate a queen-global memory save request."""
-    parsed = parse_global_memory_category(category)
-    if parsed is None:
-        raise ValueError(
-            "Invalid global memory category. Use one of: "
-            + ", ".join(GLOBAL_MEMORY_CATEGORIES)
-        )
-    if not description.strip():
-        raise ValueError("Global memory description cannot be empty.")
-    if not content.strip():
-        raise ValueError("Global memory content cannot be empty.")
-
-    probe = f"{description}\n{content}"
-    if _GLOBAL_MEMORY_CODE_PATTERN.search(probe):
-        raise ValueError(
-            "Global memory is only for durable user profile, preferences, "
-            "environment, or feedback — not task/code/runtime details."
-        )
-    return parsed
-
-
-def save_global_memory(
-    *,
-    category: str,
-    description: str,
-    content: str,
-    name: str | None = None,
-    memory_dir: Path | None = None,
-) -> tuple[str, Path]:
-    """Persist one queen-global memory entry."""
-    parsed = validate_global_memory_payload(
-        category=category,
-        description=description,
-        content=content,
-    )
-    target_dir = memory_dir or global_memory_dir()
-    target_dir.mkdir(parents=True, exist_ok=True)
-    memory_name = (name or description).strip()
-    filename = allocate_memory_filename(target_dir, memory_name)
-    doc = build_memory_document(
-        name=memory_name,
-        description=description,
-        mem_type=parsed,
-        body=content,
-    )
-    if len(doc.encode("utf-8")) > MAX_FILE_SIZE_BYTES:
-        raise ValueError(
-            f"Global memory entry exceeds the {MAX_FILE_SIZE_BYTES} byte limit."
-        )
-    path = target_dir / filename
-    path.write_text(doc, encoding="utf-8")
-    return filename, path
-
-
 # ---------------------------------------------------------------------------
 # Manifest formatting
 # ---------------------------------------------------------------------------

-def _age_label(mtime: float) -> str:
-    """Human-readable age string from an mtime."""
-    age_days = memory_age_days(mtime)
-    if age_days <= 0:
-        return "today"
-    if age_days == 1:
-        return "1 day ago"
-    return f"{age_days} days ago"
-

 def format_memory_manifest(files: list[MemoryFile]) -> str:
-    """One-line-per-file text manifest for the recall selector / reflection agent.
+    """One-line-per-file text manifest.

-    Format: ``[type] filename (age): description``
+    Format: ``[type] filename: description``
    """
    lines: list[str] = []
    for mf in files:
        t = mf.type or "unknown"
        desc = mf.description or "(no description)"
-        age = _age_label(mf.mtime)
-        lines.append(f"[{t}] {mf.filename} ({age}): {desc}")
+        lines.append(f"[{t}] {mf.filename}: {desc}")
    return "\n".join(lines)


 # ---------------------------------------------------------------------------
-# Freshness / staleness
-# ---------------------------------------------------------------------------
-
-_SECONDS_PER_DAY = 86_400
-
-
-def memory_age_days(mtime: float) -> int:
-    """Return the age of a memory file in whole days."""
-    if mtime <= 0:
-        return 0
-    return int((time.time() - mtime) / _SECONDS_PER_DAY)
-
-
-def memory_freshness_text(mtime: float) -> str:
-    """Return a staleness warning for injection, or empty string if fresh."""
-    d = memory_age_days(mtime)
-    if d <= 1:
-        return ""
-    return (
-        f"This memory is {d} days old. "
-        "Memories are point-in-time observations, not live state — "
-        "claims about code behavior or file:line citations may be outdated. "
-        "Verify against current code before asserting as fact."
-    )
-
-
-# ---------------------------------------------------------------------------
-# Cursor-based incremental processing
+# Initialisation
 # ---------------------------------------------------------------------------


-async def read_conversation_parts(session_dir: Path) -> list[dict[str, Any]]:
-    """Read all conversation parts for a session using FileConversationStore.
-
-    Returns a list of raw message dicts in sequence order.
-    """
-    from framework.storage.conversation_store import FileConversationStore
-
-    store = FileConversationStore(session_dir / "conversations")
-    return await store.read_parts()
-
-
-# ---------------------------------------------------------------------------
-# Initialisation and legacy migration
-# ---------------------------------------------------------------------------
-
-
-def init_memory_dir(
-    memory_dir: Path | None = None,
-    *,
-    migrate_legacy: bool = False,
-) -> None:
-    """Create the memory directory if missing.
-
-    When ``migrate_legacy`` is true, migrate both v1 memory files and the
-    previous shared v2 queen memory store into this directory.
-    """
-    d = memory_dir or MEMORY_DIR
-    first_run = not d.exists()
+def init_memory_dir(memory_dir: Path | None = None) -> None:
+    """Create the memory directory if missing."""
+    d = memory_dir or global_memory_dir()
    d.mkdir(parents=True, exist_ok=True)
-    if migrate_legacy:
-        migrate_legacy_memories(d)
-        migrate_shared_v2_memories(d)
-    elif first_run and d == MEMORY_DIR:
-        migrate_legacy_memories(d)
-
-
-def migrate_legacy_memories(memory_dir: Path | None = None) -> None:
-    """Convert old MEMORY.md + MEMORY-YYYY-MM-DD.md files to individual memory files.
-
-    Originals are moved to ``{memory_dir}/.legacy/``.
-    """
-    d = memory_dir or MEMORY_DIR
-    queen_dir = _HIVE_QUEEN_DIR
-    legacy_archive = d / ".legacy"
-
-    migrated_any = False
-
-    # --- Semantic memory (MEMORY.md) ---
-    semantic = queen_dir / "MEMORY.md"
-    if semantic.exists():
-        content = semantic.read_text(encoding="utf-8").strip()
-        # Skip the blank seed template.
-        if content and not content.startswith("# My Understanding of the User\n\n*No sessions"):
-            _write_migration_file(
-                d,
-                filename="legacy-semantic-memory.md",
-                name="legacy-semantic-memory",
-                mem_type="reference",
-                description="Migrated semantic memory from previous memory system",
-                body=content,
-            )
-            migrated_any = True
-        # Archive original.
-        legacy_archive.mkdir(parents=True, exist_ok=True)
-        semantic.rename(legacy_archive / "MEMORY.md")
-
-    # --- Episodic memories (MEMORY-YYYY-MM-DD.md) ---
-    old_memories_dir = queen_dir / "memories"
-    if old_memories_dir.is_dir():
-        for ep_file in sorted(old_memories_dir.glob("MEMORY-*.md")):
-            content = ep_file.read_text(encoding="utf-8").strip()
-            if not content:
-                continue
-            date_part = ep_file.stem.replace("MEMORY-", "")
-            slug = f"legacy-diary-{date_part}.md"
-            _write_migration_file(
-                d,
-                filename=slug,
-                name=f"legacy-diary-{date_part}",
-                mem_type="diary",
-                description=f"Migrated diary entry from {date_part}",
-                body=content,
-            )
-            migrated_any = True
-            # Archive original.
-            legacy_archive.mkdir(parents=True, exist_ok=True)
-            ep_file.rename(legacy_archive / ep_file.name)
-
-    if migrated_any:
-        logger.info("queen_memory_v2: migrated legacy memory files to %s", d)
-
-
-def migrate_shared_v2_memories(
-    memory_dir: Path | None = None,
-    *,
-    source_dir: Path | None = None,
-) -> None:
-    """Move shared queen v2 memory files into a colony directory once."""
-    d = memory_dir or MEMORY_DIR
-    d.mkdir(parents=True, exist_ok=True)
-    src = source_dir or MEMORY_DIR
-    if d.resolve() == src.resolve():
-        return
-
-    marker = d / _MIGRATION_MARKER
-    if marker.exists():
-        return
-
-    if not src.is_dir():
-        return
-
-    md_files = sorted(
-        f for f in src.glob("*.md")
-        if f.is_file() and not f.name.startswith(".")
-    )
-    if not md_files:
-        marker.write_text("no shared memories found\n", encoding="utf-8")
-        return
-
-    archive = src / ".legacy_colony_migration"
-    archive.mkdir(parents=True, exist_ok=True)
-    migrated_any = False
-
-    for src_file in md_files:
-        target = d / src_file.name
-        if not target.exists():
-            try:
-                shutil.copy2(src_file, target)
-                migrated_any = True
-            except OSError:
-                logger.debug("shared memory migration copy failed for %s", src_file, exc_info=True)
-                continue
-
-        archived = archive / src_file.name
-        counter = 2
-        while archived.exists():
-            archived = archive / f"{src_file.stem}-{counter}{src_file.suffix}"
-            counter += 1
-        try:
-            src_file.rename(archived)
-        except OSError:
-            logger.debug("shared memory migration archive failed for %s", src_file, exc_info=True)
-
-    if migrated_any:
-        logger.info("queen_memory_v2: migrated shared queen memories to %s", d)
-    marker.write_text(
-        f"migrated_at={int(time.time())}\nsource={src}\n",
-        encoding="utf-8",
-    )
-
-
-def _write_migration_file(
-    memory_dir: Path,
-    filename: str,
-    name: str,
-    mem_type: str,
-    description: str,
-    body: str,
-) -> None:
-    """Write a single migrated memory file with frontmatter."""
-    # Truncate body to respect file size limit (leave room for frontmatter).
-    header = (
-        f"---\n"
-        f"name: {name}\n"
-        f"description: {description}\n"
-        f"type: {mem_type}\n"
-        f"---\n\n"
-    )
-    max_body = MAX_FILE_SIZE_BYTES - len(header.encode("utf-8"))
-    if len(body.encode("utf-8")) > max_body:
-        # Rough truncation — cut at character level then trim to last newline.
-        body = body[: max_body - 20]
-        nl = body.rfind("\n")
-        if nl > 0:
-            body = body[:nl]
-        body += "\n\n...(truncated during migration)"
-
-    path = memory_dir / filename
-    path.write_text(header + body + "\n", encoding="utf-8")
@@ -0,0 +1,217 @@
+"""Per-queen tool configuration sidecar (``tools.json``).
+
+Lives at ``~/.hive/agents/queens/{queen_id}/tools.json`` alongside
+``profile.yaml``. Kept separate so identity (name, title, core traits)
+stays human-authored and lean, while the machine-managed tool allowlist
+can grow (per-tool overrides, audit timestamps, future per-phase rules)
+without bloating the profile.
+
+Schema::
+
+    {
+      "enabled_mcp_tools": ["read_file", ...] | null,
+      "updated_at": "2026-04-21T12:34:56+00:00"
+    }
+
+- ``null`` / missing file → default "allow every MCP tool".
+- ``[]`` → explicitly disable every MCP tool.
+- ``["foo", "bar"]`` → only those MCP tool names pass the filter.
+
+Atomic writes via ``os.replace`` follow the same pattern as
+``framework.host.colony_metadata.update_colony_metadata``.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import os
+import tempfile
+from datetime import UTC, datetime
+from pathlib import Path
+from typing import Any
+
+import yaml
+
+from framework.config import QUEENS_DIR
+
+logger = logging.getLogger(__name__)
+
+
+def tools_config_path(queen_id: str) -> Path:
+    """Return the on-disk path to a queen's ``tools.json``."""
+    return QUEENS_DIR / queen_id / "tools.json"
+
+
+def _atomic_write_json(path: Path, data: dict[str, Any]) -> None:
+    """Write ``data`` to ``path`` atomically via tempfile + replace."""
+    path.parent.mkdir(parents=True, exist_ok=True)
+    fd, tmp = tempfile.mkstemp(
+        prefix=".tools.",
+        suffix=".json.tmp",
+        dir=str(path.parent),
+    )
+    try:
+        with os.fdopen(fd, "w", encoding="utf-8") as fh:
+            json.dump(data, fh, indent=2)
+            fh.flush()
+            os.fsync(fh.fileno())
+        os.replace(tmp, path)
+    except BaseException:
+        try:
+            os.unlink(tmp)
+        except OSError:
+            pass
+        raise
+
+
+def _migrate_from_profile_if_needed(queen_id: str) -> list[str] | None:
+    """Hoist a legacy ``enabled_mcp_tools`` field out of ``profile.yaml``.
+
+    Returns the migrated value (or ``None`` if nothing to migrate). After
+    migration the sidecar exists on disk and the profile YAML no longer
+    contains ``enabled_mcp_tools``. Safe to call repeatedly.
+    """
+    profile_path = QUEENS_DIR / queen_id / "profile.yaml"
+    if not profile_path.exists():
+        return None
+    try:
+        data = yaml.safe_load(profile_path.read_text(encoding="utf-8"))
+    except (yaml.YAMLError, OSError):
+        logger.warning("Could not read profile.yaml during tools migration: %s", queen_id)
+        return None
+    if not isinstance(data, dict):
+        return None
+    if "enabled_mcp_tools" not in data:
+        return None
+
+    raw = data.pop("enabled_mcp_tools")
+    enabled: list[str] | None
+    if raw is None:
+        enabled = None
+    elif isinstance(raw, list) and all(isinstance(x, str) for x in raw):
+        enabled = raw
+    else:
+        logger.warning(
+            "Legacy enabled_mcp_tools on queen %s had unexpected shape %r; dropping",
+            queen_id,
+            raw,
+        )
+        enabled = None
+
+    # Write sidecar first, then rewrite profile — if the second step
+    # fails we still have the config available and won't re-migrate.
+    _atomic_write_json(
+        tools_config_path(queen_id),
+        {
+            "enabled_mcp_tools": enabled,
+            "updated_at": datetime.now(UTC).isoformat(),
+        },
+    )
+    profile_path.write_text(
+        yaml.safe_dump(data, sort_keys=False, allow_unicode=True),
+        encoding="utf-8",
+    )
+    logger.info(
+        "Migrated enabled_mcp_tools for queen %s from profile.yaml to tools.json",
+        queen_id,
+    )
+    return enabled
+
+
+def tools_config_exists(queen_id: str) -> bool:
+    """Return True when the queen has a persisted ``tools.json`` sidecar.
+
+    Used by callers that need to tell an explicit user save apart from a
+    fallthrough to the role-based default (both can return the same
+    value from ``load_queen_tools_config``).
+    """
+    return tools_config_path(queen_id).exists()
+
+
+def delete_queen_tools_config(queen_id: str) -> bool:
+    """Delete the queen's ``tools.json`` sidecar if present.
+
+    Returns ``True`` if a file was removed, ``False`` if none existed.
+    The next ``load_queen_tools_config`` call falls through to the
+    role-based default (or allow-all for unknown queens).
+    """
+    path = tools_config_path(queen_id)
+    if not path.exists():
+        return False
+    try:
+        path.unlink()
+        return True
+    except OSError:
+        logger.warning("Failed to delete %s", path, exc_info=True)
+        return False
+
+
+def load_queen_tools_config(
+    queen_id: str,
+    mcp_catalog: dict[str, list[dict]] | None = None,
+) -> list[str] | None:
+    """Return the queen's MCP tool allowlist, or ``None`` for default-allow.
+
+    Order of resolution:
+    1. ``tools.json`` sidecar (authoritative; user has saved).
+    2. Legacy ``profile.yaml`` field (migrated and deleted on first read).
+    3. Role-based default from ``queen_tools_defaults`` when the queen
+       is in the known persona table. ``mcp_catalog`` lets the helper
+       expand ``@server:NAME`` shorthands; without it, shorthand entries
+       are dropped.
+    4. ``None`` — default "allow every MCP tool".
+    """
+    path = tools_config_path(queen_id)
+    if path.exists():
+        try:
+            data = json.loads(path.read_text(encoding="utf-8"))
+        except (json.JSONDecodeError, OSError):
+            logger.warning("Invalid %s; treating as default-allow", path)
+            return None
+        if not isinstance(data, dict):
+            return None
+        raw = data.get("enabled_mcp_tools")
+        if raw is None:
+            return None
+        if isinstance(raw, list) and all(isinstance(x, str) for x in raw):
+            return raw
+        logger.warning("Unexpected enabled_mcp_tools shape in %s; ignoring", path)
+        return None
+
+    migrated = _migrate_from_profile_if_needed(queen_id)
+    if migrated is not None:
+        return migrated
+    # If migration just hoisted an explicit ``null`` out of profile.yaml,
+    # a sidecar with allow-all semantics now exists on disk. Honor that
+    # over the role default so an explicit user choice wins.
+    if tools_config_path(queen_id).exists():
+        return None
+
+    # No sidecar, nothing to migrate — fall back to role-based default.
+    from framework.agents.queen.queen_tools_defaults import resolve_queen_default_tools
+
+    return resolve_queen_default_tools(queen_id, mcp_catalog)
+
+
+def update_queen_tools_config(
+    queen_id: str,
+    enabled_mcp_tools: list[str] | None,
+) -> list[str] | None:
+    """Persist the queen's MCP allowlist to ``tools.json``.
+
+    Raises ``FileNotFoundError`` if the queen's directory is missing —
+    we refuse to silently create a sidecar for a queen that doesn't
+    exist.
+    """
+    queen_dir = QUEENS_DIR / queen_id
+    if not queen_dir.exists():
+        raise FileNotFoundError(f"Queen directory not found: {queen_id}")
+    _atomic_write_json(
+        tools_config_path(queen_id),
+        {
+            "enabled_mcp_tools": enabled_mcp_tools,
+            "updated_at": datetime.now(UTC).isoformat(),
+        },
+    )
+    return enabled_mcp_tools
@@ -0,0 +1,272 @@
+"""Role-based default tool allowlists for queens.
+
+Every queen inherits the same MCP surface (all servers loaded for the
+queen agent), but exposing 94+ tools to every persona clutters the LLM
+tool catalog and wastes prompt tokens. This module defines a sensible
+default allowlist per queen persona so, e.g., Head of Legal doesn't
+see port scanners and Head of Finance doesn't see ``apply_patch``.
+
+Defaults apply only when the queen has no ``tools.json`` sidecar — the
+moment the user saves an allowlist through the Tool Library, the
+sidecar becomes authoritative. A DELETE on the tools endpoint removes
+the sidecar and brings the queen back to her role default.
+
+Category entries support a ``@server:NAME`` shorthand that expands to
+every tool name registered against that MCP server in the current
+catalog. This keeps the category table short and drift-free when new
+tools are added (e.g. browser_* auto-joins the ``browser`` category).
+"""
+
+from __future__ import annotations
+
+import logging
+from typing import Any
+
+logger = logging.getLogger(__name__)
+
+
+# ---------------------------------------------------------------------------
+# Categories — reusable bundles of MCP tool names.
+# ---------------------------------------------------------------------------
+#
+# Each category is a flat list of either concrete tool names or the
+# ``@server:NAME`` shorthand. The shorthand expands to every tool the
+# given MCP server currently exposes (requires a live catalog; when one
+# is not available the shorthand is silently dropped so we fall back to
+# the named entries only).
+
+_TOOL_CATEGORIES: dict[str, list[str]] = {
+    # Read-only file operations — safe baseline for every knowledge queen.
+    "file_read": [
+        "read_file",
+        "list_directory",
+        "list_dir",
+        "list_files",
+        "search_files",
+        "grep_search",
+        "pdf_read",
+    ],
+    # File mutation — only personas that author or edit artifacts.
+    "file_write": [
+        "write_file",
+        "edit_file",
+        "apply_diff",
+        "apply_patch",
+        "replace_file_content",
+        "hashline_edit",
+        "undo_changes",
+    ],
+    # Shell + process control — engineering personas only.
+    "shell": [
+        "run_command",
+        "execute_command_tool",
+        "bash_kill",
+        "bash_output",
+    ],
+    # Tabular data. CSV/Excel read/write + DuckDB SQL.
+    "data": [
+        "csv_read",
+        "csv_info",
+        "csv_write",
+        "csv_append",
+        "csv_sql",
+        "excel_read",
+        "excel_info",
+        "excel_write",
+        "excel_append",
+        "excel_search",
+        "excel_sheet_list",
+        "excel_sql",
+    ],
+    # Browser automation — every tool from the gcu-tools MCP server.
+    "browser": ["@server:gcu-tools"],
+    # External research / information-gathering.
+    "research": [
+        "search_papers",
+        "download_paper",
+        "search_wikipedia",
+        "web_scrape",
+    ],
+    # Security scanners — pentest-ish, only for engineering/security roles.
+    "security": [
+        "dns_security_scan",
+        "http_headers_scan",
+        "port_scan",
+        "ssl_tls_scan",
+        "subdomain_enumerate",
+        "tech_stack_detect",
+        "risk_score",
+    ],
+    # Lightweight context helpers — good default for every queen.
+    "time_context": [
+        "get_current_time",
+        "get_account_info",
+    ],
+    # Runtime log inspection — debug/observability for builder personas.
+    "runtime_inspection": [
+        "query_runtime_logs",
+        "query_runtime_log_details",
+        "query_runtime_log_raw",
+    ],
+    # Agent-management tools — building/validating/checking agents.
+    "agent_mgmt": [
+        "list_agents",
+        "list_agent_tools",
+        "list_agent_sessions",
+        "get_agent_checkpoint",
+        "list_agent_checkpoints",
+        "run_agent_tests",
+        "save_agent_draft",
+        "confirm_and_build",
+        "validate_agent_package",
+        "validate_agent_tools",
+        "enqueue_task",
+    ],
+}
+
+
+# ---------------------------------------------------------------------------
+# Per-queen mapping.
+# ---------------------------------------------------------------------------
+#
+# Built from the queen personas in ``queen_profiles.DEFAULT_QUEENS``. The
+# goal is "just enough" — a queen should see tools she'd plausibly call
+# for her stated role, nothing more. Users curate further via the Tool
+# Library if they want.
+#
+# A queen whose ID is NOT in this map falls through to "allow every MCP
+# tool" (the original behavior), which keeps the system compatible with
+# user-added custom queen IDs that we don't know about.
+
+QUEEN_DEFAULT_CATEGORIES: dict[str, list[str]] = {
+    # Head of Technology — builds and operates systems; full toolkit.
+    "queen_technology": [
+        "file_read",
+        "file_write",
+        "shell",
+        "data",
+        "browser",
+        "research",
+        "security",
+        "time_context",
+        "runtime_inspection",
+        "agent_mgmt",
+    ],
+    # Head of Growth — data, experiments, competitor research; no shell/security.
+    "queen_growth": [
+        "file_read",
+        "file_write",
+        "data",
+        "browser",
+        "research",
+        "time_context",
+    ],
+    # Head of Product Strategy — user research + roadmaps; no shell/security.
+    "queen_product_strategy": [
+        "file_read",
+        "file_write",
+        "data",
+        "browser",
+        "research",
+        "time_context",
+    ],
+    # Head of Finance — financial models (CSV/Excel heavy), market research.
+    "queen_finance_fundraising": [
+        "file_read",
+        "file_write",
+        "data",
+        "browser",
+        "research",
+        "time_context",
+    ],
+    # Head of Legal — reads contracts/PDFs, researches; no shell/data/security.
+    "queen_legal": [
+        "file_read",
+        "file_write",
+        "browser",
+        "research",
+        "time_context",
+    ],
+    # Head of Brand & Design — visual refs, style guides; no shell/data/security.
+    "queen_brand_design": [
+        "file_read",
+        "file_write",
+        "browser",
+        "research",
+        "time_context",
+    ],
+    # Head of Talent — candidate pipelines, resumes; data + browser heavy.
+    "queen_talent": [
+        "file_read",
+        "file_write",
+        "data",
+        "browser",
+        "research",
+        "time_context",
+    ],
+    # Head of Operations — processes, automation, observability.
+    "queen_operations": [
+        "file_read",
+        "file_write",
+        "data",
+        "browser",
+        "research",
+        "time_context",
+        "runtime_inspection",
+        "agent_mgmt",
+    ],
+}
+
+
+def has_role_default(queen_id: str) -> bool:
+    """Return True when ``queen_id`` is known to the category table."""
+    return queen_id in QUEEN_DEFAULT_CATEGORIES
+
+
+def resolve_queen_default_tools(
+    queen_id: str,
+    mcp_catalog: dict[str, list[dict[str, Any]]] | None = None,
+) -> list[str] | None:
+    """Return the role-based default allowlist for ``queen_id``.
+
+    Arguments:
+        queen_id: Profile ID (e.g. ``"queen_technology"``).
+        mcp_catalog: Optional mapping of ``{server_name: [{"name": ...}, ...]}``
+            used to expand ``@server:NAME`` shorthands in categories.
+            When absent, shorthand entries are dropped and the result
+            contains only the explicitly-named tools.
+
+    Returns:
+        A deduplicated list of tool names, or ``None`` if the queen has
+        no role entry (caller should treat as "allow every MCP tool").
+    """
+    categories = QUEEN_DEFAULT_CATEGORIES.get(queen_id)
+    if not categories:
+        return None
+
+    names: list[str] = []
+    seen: set[str] = set()
+
+    def _add(name: str) -> None:
+        if name and name not in seen:
+            seen.add(name)
+            names.append(name)
+
+    for cat in categories:
+        for entry in _TOOL_CATEGORIES.get(cat, []):
+            if entry.startswith("@server:"):
+                server_name = entry[len("@server:") :]
+                if mcp_catalog is None:
+                    logger.debug(
+                        "resolve_queen_default_tools: catalog missing; cannot expand %s",
+                        entry,
+                    )
+                    continue
+                for tool in mcp_catalog.get(server_name, []) or []:
+                    tname = tool.get("name") if isinstance(tool, dict) else None
+                    if tname:
+                        _add(tname)
+            else:
+                _add(entry)
+
+    return names
@@ -1,11 +1,11 @@
-"""Recall selector — pre-turn memory selection for queen and worker memory.
+"""Recall selector — pre-turn memory selection for the queen.

 Before each conversation turn the system:
-  1. Scans the memory directory for ``.md`` files (cap: 200).
+  1. Scans one or more memory directories for ``.md`` files (cap: 200 each).
  2. Reads headers (frontmatter + first 30 lines).
-  3. Uses a single LLM call with structured JSON output to pick the ~5
-     most relevant memories.
-  4. Injects them into context with staleness warnings for older ones.
+  3. Uses an LLM call with structured JSON output to pick the most relevant
+     memories for each scope.
+  4. Injects them into the system prompt.

 The selector only sees the user's query string — no full conversation
 context.  This keeps it cheap and fast.  Errors are caught and return
@@ -20,9 +20,8 @@ from pathlib import Path
 from typing import Any

 from framework.agents.queen.queen_memory_v2 import (
-    MEMORY_DIR,
    format_memory_manifest,
-    memory_freshness_text,
+    global_memory_dir as _default_global_memory_dir,
    scan_memory_files,
 )

@@ -32,29 +31,6 @@ logger = logging.getLogger(__name__)
 # Structured output schema
 # ---------------------------------------------------------------------------

-RECALL_SCHEMA: dict[str, Any] = {
-    "type": "json_schema",
-    "json_schema": {
-        "name": "memory_selection",
-        "strict": True,
-        "schema": {
-            "type": "object",
-            "properties": {
-                "selected_memories": {
-                    "type": "array",
-                    "items": {"type": "string"},
-                },
-            },
-            "required": ["selected_memories"],
-            "additionalProperties": False,
-        },
-    },
-}
-
-# ---------------------------------------------------------------------------
-# System prompt
-# ---------------------------------------------------------------------------
-
 SELECT_MEMORIES_SYSTEM_PROMPT = """\
 You are selecting memories that will be useful to the Queen agent as it \
 processes a user's query.
@@ -72,9 +48,6 @@ name and description.
 query, then do not include it in your list.  Be selective and discerning.
 - If there are no memories in the list that would clearly be useful, \
 return an empty list.
- If a list of recently-used tools is provided, do not select memories \
-that are usage reference or API documentation for those tools (the Queen \
-is already exercising them).  Still select warnings or gotchas about them.
 """

 # ---------------------------------------------------------------------------
@@ -86,7 +59,6 @@ async def select_memories(
    query: str,
    llm: Any,
    memory_dir: Path | None = None,
-    active_tools: list[str] | None = None,
    *,
    max_results: int = 5,
 ) -> list[str]:
@@ -94,51 +66,83 @@ async def select_memories(

    Returns a list of filenames.  Best-effort: on any error returns ``[]``.
    """
-    mem_dir = memory_dir or MEMORY_DIR
+    mem_dir = memory_dir or _default_global_memory_dir()
    files = scan_memory_files(mem_dir)
    if not files:
        logger.debug("recall: no memory files found, skipping selection")
        return []

-    logger.debug("recall: selecting from %d memory files for query: %.80s", len(files), query)
+    logger.debug("recall: selecting from %d memories for query: %.100s", len(files), query)
    manifest = format_memory_manifest(files)
-
-    user_msg_parts = [f"## User query\n\n{query}\n\n## Available memories\n\n{manifest}"]
-    if active_tools:
-        user_msg_parts.append(f"\n\n## Recently-used tools\n\n{', '.join(active_tools)}")
-
-    user_msg = "".join(user_msg_parts)
+    user_msg = f"## User query\n\n{query}\n\n## Available memories\n\n{manifest}"

    try:
        resp = await llm.acomplete(
            messages=[{"role": "user", "content": user_msg}],
            system=SELECT_MEMORIES_SYSTEM_PROMPT,
-            max_tokens=512,
-            response_format=RECALL_SCHEMA,
+            max_tokens=1024,
+            response_format={"type": "json_object"},
        )
-        data = json.loads(resp.content)
+        raw = (resp.content or "").strip()
+        if not raw:
+            logger.warning(
+                "recall: LLM returned empty response (model=%s, stop=%s)",
+                resp.model,
+                resp.stop_reason,
+            )
+            return []
+        # Some models wrap JSON in markdown fences or add preamble text.
+        # Try to extract the JSON object if raw parse fails.
+        try:
+            data = json.loads(raw)
+        except json.JSONDecodeError:
+            import re
+
+            m = re.search(r"\{.*\}", raw, re.DOTALL)
+            if m:
+                data = json.loads(m.group())
+            else:
+                logger.warning("recall: LLM returned non-JSON: %.200s", raw)
+                return []
        selected = data.get("selected_memories", [])
-        # Validate: only return filenames that actually exist.
        valid_names = {f.filename for f in files}
        result = [s for s in selected if s in valid_names][:max_results]
        logger.debug("recall: selected %d memories: %s", len(result), result)
        return result
-    except Exception:
-        logger.debug("recall: memory selection failed, returning []", exc_info=True)
+    except Exception as exc:
+        logger.warning("recall: memory selection failed (%s), returning []", exc)
        return []


+def _format_relative_age(mtime: float) -> str | None:
+    """Return age description if memory is older than 48 hours.
+
+    Returns None if 48 hours or newer, otherwise returns "X days old".
+    """
+    import time
+
+    age_seconds = time.time() - mtime
+    hours = age_seconds / 3600
+    if hours <= 48:
+        return None
+    days = int(age_seconds / 86400)
+    if days == 1:
+        return "1 day old"
+    return f"{days} days old"
+
+
 def format_recall_injection(
    filenames: list[str],
    memory_dir: Path | None = None,
    *,
-    heading: str = "Selected Memories",
+    label: str = "Global Memories",
 ) -> str:
    """Read selected memory files and format for system prompt injection.

-    Prepends a staleness warning for memories older than 1 day.
+    Includes relative timestamp (e.g., "3 days old") for memories older than 48 hours.
    """
-    mem_dir = memory_dir or MEMORY_DIR
+
+    mem_dir = memory_dir or _default_global_memory_dir()
    if not filenames:
        return ""

@@ -149,88 +153,63 @@ def format_recall_injection(
            continue
        try:
            content = path.read_text(encoding="utf-8").strip()
+            # Get file modification time for age calculation
+            mtime = path.stat().st_mtime
+            age_note = _format_relative_age(mtime)
        except OSError:
            continue

-        try:
-            mtime = path.stat().st_mtime
-        except OSError:
-            mtime = 0.0
-
-        freshness = memory_freshness_text(mtime)
-        header = f"### {fname}"
-        if freshness:
-            header += f"\n\n> {freshness}"
+        # Build header with optional age note
+        if age_note:
+            header = f"### {fname} ({age_note})"
+        else:
+            header = f"### {fname}"
        blocks.append(f"{header}\n\n{content}")

    if not blocks:
        return ""

    body = "\n\n---\n\n".join(blocks)
-    logger.debug("recall: injecting %d memory blocks into context", len(blocks))
-    return f"--- {heading} ---\n\n{body}\n\n--- End {heading} ---"
+    return f"--- {label} ---\n\n{body}\n\n--- End {label} ---"


-# ---------------------------------------------------------------------------
-# Cache update (called after each queen turn)
-# ---------------------------------------------------------------------------
-
-
-async def update_recall_cache(
-    session_dir: Path,
+async def build_scoped_recall_blocks(
+    query: str,
    llm: Any,
-    phase_state: Any | None = None,
-    memory_dir: Path | None = None,
    *,
-    cache_setter: Any = None,
-    heading: str = "Selected Memories",
-    active_tools: list[str] | None = None,
-) -> None:
-    """Update the recall cache on *phase_state* for the next turn.
+    global_memory_dir: Path | None = None,
+    queen_memory_dir: Path | None = None,
+    queen_id: str | None = None,
+    global_max_results: int = 3,
+    queen_max_results: int = 3,
+) -> tuple[str, str]:
+    """Build separate recall blocks for global and queen-scoped memory."""
+    global_dir = global_memory_dir or _default_global_memory_dir()
+    global_selected = await select_memories(
+        query,
+        llm,
+        memory_dir=global_dir,
+        max_results=global_max_results,
+    )
+    global_block = format_recall_injection(
+        global_selected,
+        memory_dir=global_dir,
+        label="Global Memories",
+    )

-    Reads the latest user message from conversation parts to use as the
-    query for memory selection.
-    """
-    mem_dir = memory_dir or MEMORY_DIR
-
-    # Extract latest user message as the query.
-    query = _extract_latest_user_query(session_dir)
-    if not query:
-        logger.debug("recall: no user query found, skipping cache update")
-        return
-    logger.debug("recall: updating cache for query: %.80s", query)
-
-    try:
-        selected = await select_memories(
+    queen_block = ""
+    if queen_memory_dir is not None:
+        queen_selected = await select_memories(
            query,
            llm,
-            mem_dir,
-            active_tools=active_tools,
+            memory_dir=queen_memory_dir,
+            max_results=queen_max_results,
+        )
+        queen_label = f"Queen Memories: {queen_id}" if queen_id else "Queen Memories"
+        queen_block = format_recall_injection(
+            queen_selected,
+            memory_dir=queen_memory_dir,
+            label=queen_label,
        )
-        injection = format_recall_injection(selected, mem_dir, heading=heading)
-        if cache_setter is not None:
-            cache_setter(injection)
-        elif phase_state is not None:
-            phase_state._cached_recall_block = injection
-    except Exception:
-        logger.debug("recall: cache update failed", exc_info=True)

-
-def _extract_latest_user_query(session_dir: Path) -> str:
-    """Read the most recent user message from conversation parts."""
-    parts_dir = session_dir / "conversations" / "parts"
-    if not parts_dir.is_dir():
-        return ""
-
-    part_files = sorted(parts_dir.glob("*.json"), reverse=True)
-    for f in part_files[:20]:  # Look back at most 20 messages.
-        try:
-            data = json.loads(f.read_text(encoding="utf-8"))
-            if data.get("role") == "user":
-                content = str(data.get("content", "")).strip()
-                if content:
-                    # Truncate very long queries.
-                    return content[:1000] if len(content) > 1000 else content
-        except (json.JSONDecodeError, OSError):
-            continue
-    return ""
+    return global_block, queen_block
@@ -13,7 +13,7 @@
 6. **Calling set_output in same turn as tool calls** — Call set_output in a SEPARATE turn.

 ## File Template Errors
-7. **Wrong import paths** — Use `from framework.graph import ...`, NOT `from core.framework.graph import ...`.
+7. **Wrong import paths** — Use `from framework.orchestrator import ...`, NOT `from framework.graph import ...` or `from core.framework...`.
 8. **Missing storage path** — Agent class must set `self._storage_path = Path.home() / ".hive" / "agents" / "agent_name"`.
 9. **Missing mcp_servers.json** — Without this, the agent has no tools at runtime.
 10. **Bare `python` command** — Use `"command": "uv"` with args `["run", "python", ...]`.
@@ -25,10 +25,7 @@
 14. **Forgetting sys.path setup in conftest.py** — Tests need `exports/` and `core/` on sys.path.

 ## GCU Errors
-15. **Manually wiring browser tools on event_loop nodes** — Use `node_type="gcu"` which auto-includes browser tools. Do NOT manually list browser tool names.
-16. **Using GCU nodes as regular graph nodes** — GCU nodes are subagents only. They must ONLY appear in `sub_agents=["gcu-node-id"]` and be invoked via `delegate_to_sub_agent()`. Never connect via edges or use as entry/terminal nodes.
-17. **Reusing the same GCU node ID for parallel tasks** — Each concurrent browser task needs a distinct GCU node ID (e.g. `gcu-site-a`, `gcu-site-b`). Two `delegate_to_sub_agent` calls with the same `agent_id` share a browser profile and will interfere with each other's pages.
-18. **Passing `profile=` in GCU tool calls** — Profile isolation for parallel subagents is automatic. The framework injects a unique profile per subagent via an asyncio `ContextVar`. Hardcoding `profile="default"` in a GCU system prompt breaks this isolation.
+15. **Manually wiring browser tools on event_loop nodes** — Browser nodes use tools: {policy: "all"} to get all browser tools.

 ## Worker Agent Errors
 19. **Adding client-facing intake node to workers** — The queen owns intake. Workers should start with an autonomous processing node. Route worker review/approval through queen escalation instead of direct worker HITL.
@@ -55,7 +55,7 @@ metadata = AgentMetadata()
 ```python
 """Node definitions for My Agent."""

-from framework.graph import NodeSpec
+from framework.orchestrator import NodeSpec

 # Node 1: Process (autonomous entry node)
 # The queen handles intake and passes structured input via
@@ -123,14 +123,15 @@ __all__ = ["process_node", "handoff_node"]

 from pathlib import Path

-from framework.graph import EdgeSpec, EdgeCondition, Goal, SuccessCriterion, Constraint
-from framework.graph.edge import GraphSpec
-from framework.graph.executor import ExecutionResult
-from framework.graph.checkpoint_config import CheckpointConfig
+from framework.orchestrator import EdgeSpec, EdgeCondition, Goal, SuccessCriterion, Constraint
+from framework.orchestrator.edge import GraphSpec
+from framework.orchestrator.orchestrator import ExecutionResult
+from framework.orchestrator.checkpoint_config import CheckpointConfig
 from framework.llm import LiteLLMProvider
-from framework.runner.tool_registry import ToolRegistry
-from framework.runtime.agent_runtime import AgentRuntime, create_agent_runtime
-from framework.runtime.execution_stream import EntryPointSpec
+from framework.loader.tool_registry import ToolRegistry
+from framework.host.agent_host import AgentHost
+from framework.host.execution_manager import EntryPointSpec
+

 from .config import default_config, metadata
 from .nodes import process_node, handoff_node
@@ -227,7 +228,7 @@ class MyAgent:
        tools = list(self._tool_registry.get_tools().values())
        tool_executor = self._tool_registry.get_executor()
        self._graph = self._build_graph()
-        self._agent_runtime = create_agent_runtime(
+        self._agent_runtime = AgentHost(
            graph=self._graph, goal=self.goal, storage_path=self._storage_path,
            entry_points=[EntryPointSpec(id="default", name="Default", entry_node=self.entry_node,
                                         trigger_type="manual", isolation_level="shared")],
@@ -460,8 +461,8 @@ def tui():
    from framework.tui.app import AdenTUI
    from framework.llm import LiteLLMProvider
    from framework.runner.tool_registry import ToolRegistry
-    from framework.runtime.agent_runtime import create_agent_runtime
-    from framework.runtime.execution_stream import EntryPointSpec
+    from framework.host.agent_host import AgentHost
+    from framework.host.execution_manager import EntryPointSpec

    async def run_tui():
        agent = MyAgent()
@@ -471,7 +472,7 @@ def tui():
        mcp_cfg = Path(__file__).parent / "mcp_servers.json"
        if mcp_cfg.exists(): agent._tool_registry.load_mcp_config(mcp_cfg)
        llm = LiteLLMProvider(model=agent.config.model, api_key=agent.config.api_key, api_base=agent.config.api_base)
-        runtime = create_agent_runtime(
+        runtime = AgentHost(
            graph=agent._build_graph(), goal=agent.goal, storage_path=storage,
            entry_points=[EntryPointSpec(id="start", name="Start", entry_node="process", trigger_type="manual", isolation_level="isolated")],
            llm=llm, tools=list(agent._tool_registry.get_tools().values()), tool_executor=agent._tool_registry.get_executor())
@@ -509,17 +510,17 @@ if __name__ == "__main__":

 ## mcp_servers.json

-> **Auto-generated.** `initialize_and_build_agent` creates this file with hive-tools
+> **Auto-generated.** `initialize_and_build_agent` creates this file with hive_tools
 > as the default. Only edit manually to add additional MCP servers.

 ```json
 {
-  "hive-tools": {
+  "hive_tools": {
    "transport": "stdio",
    "command": "uv",
    "args": ["run", "python", "mcp_server.py", "--stdio"],
    "cwd": "../../tools",
-    "description": "Hive tools MCP server"
+    "description": "hive_tools MCP server"
  }
 }
 ```
@@ -0,0 +1,227 @@
+# Declarative Agent File Templates
+
+Agents are defined as a single `agent.yaml` file. No Python code needed.
+The runner loads this file directly -- no `agent.py`, `config.py`, or
+`nodes/__init__.py` required.
+
+## agent.yaml -- Complete Agent Definition
+
+```yaml
+name: my-agent
+version: 1.0.0
+description: What this agent does.
+
+metadata:
+  intro_message: Welcome! What would you like me to do?
+
+# Template variables -- substituted into system_prompt and identity_prompt
+# via {{variable_name}} syntax.  Use this for config values that appear
+# in prompts (spreadsheet IDs, API endpoints, account names, etc.)
+variables:
+  spreadsheet_id: "1ZVxWDL..."
+  sheet_name: "contacts"
+
+goal:
+  description: What this agent achieves.
+  success_criteria:
+    - "First success criterion"
+    - "Second success criterion"
+  constraints:
+    - "Hard constraint the agent must respect"
+
+identity_prompt: |
+  You are a helpful agent.
+
+conversation_mode: continuous   # always "continuous" for Hive agents
+
+loop_config:
+  max_iterations: 100
+  max_tool_calls_per_turn: 30
+  max_context_tokens: 32000
+
+# MCP servers to connect (resolved by name from ~/.hive/mcp_registry/)
+mcp_servers:
+  - name: hive_tools
+  - name: gcu-tools
+
+nodes:
+  # Node 1: Process (autonomous entry node)
+  # The queen handles intake and passes structured input via
+  # run_agent_with_input(task). NO client-facing intake node.
+  - id: process
+    name: Process
+    description: Execute the task using available tools
+    max_node_visits: 0   # 0 = unlimited (forever-alive agents)
+    input_keys: [user_request, feedback]
+    output_keys: [results]
+    nullable_output_keys: [feedback]
+    tools:
+      policy: explicit
+      allowed: [web_search, web_scrape, save_data, load_data, list_data_files]
+    success_criteria: Results are complete and accurate.
+    system_prompt: |
+      You are a processing agent. Your task is in memory under "user_request".
+      If "feedback" is present, this is a revision.
+
+      Work in phases:
+      1. Use tools to gather/process data
+      2. Analyze results
+      3. Call set_output in a SEPARATE turn:
+         - set_output("results", "structured results")
+
+  # Node 2: Handoff (autonomous)
+  - id: handoff
+    name: Handoff
+    description: Prepare worker results for queen review
+    max_node_visits: 0
+    input_keys: [results, user_request]
+    output_keys: [next_action, feedback, worker_summary]
+    nullable_output_keys: [feedback, worker_summary]
+    tools:
+      policy: none   # handoff nodes don't need tools
+    success_criteria: Results are packaged for queen decision-making.
+    system_prompt: |
+      Do NOT talk to the user directly. The queen is the only user interface.
+
+      If blocked, call escalate(reason, context) then set:
+      - set_output("next_action", "escalated")
+      - set_output("feedback", "what help is needed")
+
+      Otherwise summarize and set:
+      - set_output("worker_summary", "short summary for queen")
+      - set_output("next_action", "done") or "revise"
+      - set_output("feedback", "what to revise") only when revising
+
+edges:
+  - from_node: process
+    to_node: handoff
+  # Feedback loop
+  - from_node: handoff
+    to_node: process
+    condition: conditional
+    condition_expr: "str(next_action).lower() == 'revise'"
+    priority: 2
+  # Escalation loop
+  - from_node: handoff
+    to_node: process
+    condition: conditional
+    condition_expr: "str(next_action).lower() == 'escalated'"
+    priority: 3
+  # Loop back for next task
+  - from_node: handoff
+    to_node: process
+    condition: conditional
+    condition_expr: "str(next_action).lower() == 'done'"
+
+entry_node: process
+terminal_nodes: []   # [] = forever-alive
+```
+
+## Key differences from Python templates
+
+| Before (Python)                     | After (YAML)                           |
+|-------------------------------------|----------------------------------------|
+| `agent.py` (250 lines boilerplate)  | Not needed                             |
+| `config.py` (dataclass + metadata)  | `variables:` + `metadata:` in YAML     |
+| `nodes/__init__.py` (NodeSpec calls)| `nodes:` list in YAML                  |
+| `__init__.py`, `__main__.py`        | Not needed                             |
+| f-string config injection           | `{{variable_name}}` templates          |
+| `mcp_servers.json` (separate file)  | `mcp_servers:` in YAML (or keep file)  |
+
+## Node types
+
+| Type         | Description                           | Tools                    |
+|--------------|---------------------------------------|--------------------------|
+| `event_loop` | LLM-driven orchestration (default)    | Explicit list or `none`  |
+| `gcu`        | Browser automation via GCU tools      | `policy: all` (auto)     |
+
+## Tool access policies
+
+```yaml
+# Explicit list (recommended for most nodes)
+tools:
+  policy: explicit
+  allowed: [web_search, save_data]
+
+# All tools (for browser automation nodes)
+tools:
+  policy: all
+
+# No tools (for handoff/summary nodes)
+tools:
+  policy: none
+```
+
+## Edge conditions
+
+| Condition     | When to use                                           |
+|---------------|-------------------------------------------------------|
+| `on_success`  | Default. Next node after current succeeds.            |
+| `on_failure`  | Fallback path when current node fails.                |
+| `always`      | Always traverse regardless of outcome.                |
+| `conditional` | Evaluate `condition_expr` against shared memory keys. |
+| `llm_decide`  | Let the LLM decide at runtime.                        |
+
+## Template variables
+
+Use `{{variable_name}}` in `system_prompt` and `identity_prompt`.
+Variables are defined in the top-level `variables:` map.
+
+```yaml
+variables:
+  spreadsheet_id: "1ZVxWDL..."
+  api_endpoint: "https://api.example.com"
+
+nodes:
+  - id: start
+    system_prompt: |
+      Connect to spreadsheet: {{spreadsheet_id}}
+      API endpoint: {{api_endpoint}}
+```
+
+## Entry points
+
+Default is a single manual entry point. For timer/scheduled triggers:
+
+```yaml
+entry_points:
+  - id: default
+    trigger_type: manual
+  - id: daily-check
+    trigger_type: timer
+    trigger_config:
+      interval_minutes: 30
+```
+
+## mcp_servers.json -- Still Supported
+
+The `mcp_servers.json` file is still loaded automatically if present alongside
+`agent.yaml`.  You can also inline servers in the YAML:
+
+```yaml
+mcp_servers:
+  - name: hive_tools
+  - name: gcu-tools
+```
+
+Both approaches work. The JSON file takes precedence for backward compatibility.
+
+## Migration from Python agents
+
+Run the migration tool to convert existing agents:
+
+```bash
+uv run python -m framework.tools.migrate_agent exports/my_agent
+```
+
+This generates `agent.yaml` from the existing `agent.py` + `nodes/` + `config.py`.
+The original files are left untouched. Once verified, you can delete the Python files.
+
+## Files after migration
+
+```
+my_agent/
+  agent.yaml           # The only required file
+  mcp_servers.json     # Optional (can inline in YAML)
+  flowchart.json       # Optional (auto-generated)
+```
@@ -1,306 +1,193 @@
-# Hive Agent Framework — Condensed Reference
+# Hive Agent Framework -- Condensed Reference

 ## Architecture

-Agents are Python packages in `exports/`:
+Agents are declarative JSON configs in `exports/`:
 ```
 exports/my_agent/
-├── __init__.py          # MUST re-export ALL module-level vars from agent.py
-├── __main__.py          # CLI (run, tui, info, validate, shell)
-├── agent.py             # Graph construction (goal, edges, agent class)
-├── config.py            # Runtime config
-├── nodes/__init__.py    # Node definitions (NodeSpec)
-├── mcp_servers.json     # MCP tool server config
-└── tests/               # pytest tests
+  agent.json          # The entire agent definition
+  mcp_servers.json    # MCP tool server config (optional, prefer registry refs)
 ```

-## Agent Loading Contract
+No Python files. No `__init__.py`, `__main__.py`, `config.py`, or `nodes/`.

-`AgentRunner.load()` imports the package (`__init__.py`) and reads these
-module-level variables via `getattr()`:
+## Agent Loading

-| Variable | Required | Default if missing | Consequence |
-|----------|----------|--------------------|-------------|
-| `goal` | YES | `None` | **FATAL** — "must define goal, nodes, edges" |
-| `nodes` | YES | `None` | **FATAL** — same error |
-| `edges` | YES | `None` | **FATAL** — same error |
-| `entry_node` | no | `nodes[0].id` | Probably wrong node |
-| `entry_points` | no | `{}` | **Nodes unreachable** — validation fails |
-| `terminal_nodes` | **YES** | `[]` | **FATAL** — graph must have at least one terminal node |
-| `pause_nodes` | no | `[]` | OK |
-| `conversation_mode` | no | not passed | Isolated mode (no context carryover) |
-| `identity_prompt` | no | not passed | No agent-level identity |
-| `loop_config` | no | `{}` | No iteration limits |
-| `triggers.json` (file) | no | not present | No triggers (timers, webhooks) |
+`AgentLoader.load()` reads `agent.json` and builds the execution graph.
+If `agent.py` exists (legacy), it's loaded as a Python module instead.

-**CRITICAL:** `__init__.py` MUST import and re-export ALL of these from
-`agent.py`. Missing exports silently fall back to defaults, causing
-hard-to-debug failures.
+## agent.json Schema

-**Why `default_agent.validate()` is NOT sufficient:**
-`validate()` checks the agent CLASS's internal graph (self.nodes, self.edges).
-These are always correct because the constructor references agent.py's module
-vars directly. But `AgentRunner.load()` reads from the PACKAGE (`__init__.py`),
-not the class. So `validate()` passes while `AgentRunner.load()` fails.
-Always test with `AgentRunner.load("exports/{name}")` — this is the same
-code path the TUI and `hive run` use.
-
-## Goal
-
-Defines success criteria and constraints:
-```python
-goal = Goal(
-    id="kebab-case-id",
-    name="Display Name",
-    description="What the agent does",
-    success_criteria=[
-        SuccessCriterion(id="sc-id", description="...", metric="...", target="...", weight=0.25),
-    ],
-    constraints=[
-        Constraint(id="c-id", description="...", constraint_type="hard", category="quality"),
-    ],
-)
+```json
+{
+  "name": "my-agent",
+  "version": "1.0.0",
+  "description": "What this agent does",
+  "goal": {
+    "description": "What to achieve",
+    "success_criteria": ["criterion 1", "criterion 2"],
+    "constraints": ["constraint 1"]
+  },
+  "identity_prompt": "You are a helpful agent.",
+  "conversation_mode": "continuous",
+  "loop_config": {
+    "max_iterations": 100,
+    "max_tool_calls_per_turn": 30,
+    "max_context_tokens": 32000
+  },
+  "mcp_servers": [
+    {"name": "hive_tools"},
+    {"name": "gcu-tools"}
+  ],
+  "variables": {
+    "spreadsheet_id": "1ZVx..."
+  },
+  "nodes": [...],
+  "edges": [...],
+  "entry_node": "process",
+  "terminal_nodes": []
+}
 ```
- 3-5 success criteria, weights sum to 1.0
- 1-5 constraints (hard/soft, categories: quality, accuracy, interaction, functional)

-## NodeSpec Fields
+## Template Variables
+
+Use `{{variable_name}}` in `system_prompt` and `identity_prompt`. Variables
+are defined in the top-level `variables` object:
+
+```json
+{
+  "variables": {"sheet_id": "1ZVx..."},
+  "nodes": [{
+    "id": "start",
+    "system_prompt": "Use sheet: {{sheet_id}}"
+  }]
+}
+```
+
+## Node Fields

 | Field | Type | Default | Description |
 |-------|------|---------|-------------|
 | id | str | required | kebab-case identifier |
-| name | str | required | Display name |
+| name | str | id | Display name |
 | description | str | required | What the node does |
-| node_type | str | required | `"event_loop"` or `"gcu"` (browser automation — see GCU Guide appendix) |
-| input_keys | list[str] | required | Memory keys this node reads |
-| output_keys | list[str] | required | Memory keys this node writes via set_output |
+| node_type | str | "event_loop" | `"event_loop"` |
+| input_keys | list | [] | Memory keys this node reads |
+| output_keys | list | [] | Memory keys this node writes via set_output |
 | system_prompt | str | "" | LLM instructions |
-| tools | list[str] | [] | Tool names from MCP servers |
-| client_facing | bool | False | Deprecated compatibility field. Queen interactivity is implicit; workers should escalate instead |
-| nullable_output_keys | list[str] | [] | Keys that may remain unset |
-| max_node_visits | int | 0 | 0=unlimited (default); >1 for one-shot feedback loops |
-| max_retries | int | 3 | Retries on failure |
+| tools | object | {} | Tool access policy (see below) |
+| nullable_output_keys | list | [] | Keys that may remain unset |
+| max_node_visits | int | 1 | 0=unlimited (for forever-alive agents) |
 | success_criteria | str | "" | Natural language for judge evaluation |
+| client_facing | bool | false | Whether output is shown to user |

-## EdgeSpec Fields
+## Tool Access Policies
+
+Each node declares its tools via a policy object:
+
+```json
+{"tools": {"policy": "explicit", "allowed": ["web_search", "save_data"]}}
+{"tools": {"policy": "all"}}
+{"tools": {"policy": "none"}}
+```
+
+- `explicit` (default): only named tools. Empty `allowed` = zero tools.
+- `all`: all tools from registry (e.g. for browser automation nodes).
+- `none`: no tools (for handoff/summary nodes).
+
+## Edge Fields

 | Field | Type | Description |
 |-------|------|-------------|
-| id | str | kebab-case identifier |
-| source | str | Source node ID |
-| target | str | Target node ID |
-| condition | EdgeCondition | ON_SUCCESS, ON_FAILURE, ALWAYS, CONDITIONAL |
-| condition_expr | str | Python expression evaluated against memory (for CONDITIONAL) |
-| priority | int | Positive=forward (evaluated first), negative=feedback (loop-back) |
+| from_node | str | Source node ID |
+| to_node | str | Target node ID |
+| condition | str | `on_success`, `on_failure`, `always`, `conditional` |
+| condition_expr | str | Python expression for conditional routing |
+| priority | int | Higher = evaluated first |
+
+condition_expr examples:
+- `"needs_more_research == True"`
+- `"str(next_action).lower() == 'revise'"`

 ## Key Patterns

-### STEP 1/STEP 2 (Client-Facing Nodes)
-```
-**STEP 1 — Respond to the user (text only, NO tool calls):**
-[Present information, ask questions]
-
-**STEP 2 — After the user responds, call set_output:**
- set_output("key", "value based on user response")
-```
-This prevents premature set_output before user interaction.
-
 ### Fewer, Richer Nodes (CRITICAL)

-**Hard limit: 3-6 nodes for most agents.** Never exceed 6 unless the user
-explicitly requests a complex multi-phase pipeline.
+**Hard limit: 3-6 nodes for most agents.** Each node boundary serializes
+outputs and destroys in-context information. Merge unless:
+1. Client-facing boundary (different interaction models)
+2. Disjoint tool sets
+3. Parallel execution (fan-out branches)

-Each node boundary serializes outputs to the shared buffer and **destroys** all
-in-context information: tool call results, intermediate reasoning, conversation
-history. A research node that searches, fetches, and analyzes in ONE node keeps
-all source material in its conversation context. Split across 3 nodes, each
-downstream node only sees the serialized summary string.
-
-**Decision framework — merge unless ANY of these apply:**
-1. **Client-facing boundary** — Autonomous and client-facing work MUST be
-   separate nodes (different interaction models)
-2. **Disjoint tool sets** — If tools are fundamentally different (e.g., web
-   search vs database), separate nodes make sense
-3. **Parallel execution** — Fan-out branches must be separate nodes
-
-**Red flags that you have too many nodes:**
- A node with 0 tools (pure LLM reasoning) → merge into predecessor/successor
- A node that sets only 1 trivial output → collapse into predecessor
- Multiple consecutive autonomous nodes → combine into one rich node
- A "report" node that presents analysis → merge into the client-facing node
- A "confirm" or "schedule" node that doesn't call any external service → remove
-
-**Typical agent structure (2 nodes):**
+**Typical structure (2 nodes):**
 ```
-process (autonomous) ←→ review (queen-mediated)
-```
-The queen owns intake — she gathers requirements from the user, then
-passes structured input via `run_agent_with_input(task)`. When building
-the agent, design the entry node's `input_keys` to match what the queen
-will provide at run time. Worker agents should NOT have a client-facing
-intake node. Mid-execution review/approval should happen through queen
-escalation rather than direct worker HITL.
-
-For simpler agents, just 1 autonomous node:
-```
-process (autonomous) — loops back to itself
+process (autonomous) <-> review (queen-mediated)
 ```

-### nullable_output_keys
-For inputs that only arrive on certain edges:
-```python
-research_node = NodeSpec(
-    input_keys=["brief", "feedback"],
-    nullable_output_keys=["feedback"],  # Only present on feedback edge
-    max_node_visits=3,
-)
-```
-
-### Mutually Exclusive Outputs
-For routing decisions:
-```python
-review_node = NodeSpec(
-    output_keys=["approved", "feedback"],
-    nullable_output_keys=["approved", "feedback"],  # Node sets one or the other
-)
-```
-
-### Continuous Loop Pattern
-Mark the primary event_loop node as terminal: `terminal_nodes=["process"]`.
-The node has `output_keys` and can complete when the agent finishes its work.
-Use `conversation_mode="continuous"` to preserve context across transitions.
+The queen owns intake. Worker agents should NOT have a client-facing intake
+node. Mid-execution review should happen through queen escalation.

 ### set_output
 - Synthetic tool injected by framework
 - Call separately from real tool calls (separate turn)
 - `set_output("key", "value")` stores to the shared buffer

-## Edge Conditions
-
-| Condition | When |
-|-----------|------|
-| ON_SUCCESS | Node completed successfully |
-| ON_FAILURE | Node failed |
-| ALWAYS | Unconditional |
-| CONDITIONAL | condition_expr evaluates to True against memory |
-
-condition_expr examples:
- `"needs_more_research == True"`
- `"str(next_action).lower() == 'new_agent'"`
- `"feedback is not None"`
-
-## Graph Lifecycle
+### Graph Lifecycle

 | Pattern | terminal_nodes | When |
 |---------|---------------|------|
-| **Continuous loop** | `["node-with-output-keys"]` | **DEFAULT for all agents** |
+| Continuous loop | `["node-with-output-keys"]` | DEFAULT for all agents |
 | Linear | `["last-node"]` | One-shot/batch agents |

-**Every graph must have at least one terminal node.** Terminal nodes
-define where execution ends. For interactive agents that loop continuously,
-mark the primary event_loop node as terminal (it has `output_keys` and can
-complete at any point). The framework default for `max_node_visits` is 0
-(unbounded), so nodes work correctly in continuous loops without explicit
-override. Only set `max_node_visits > 0` in one-shot agents with feedback loops.
-Every node must have at least one outgoing edge — no dead ends.
+Every graph must have at least one terminal node.

-## Continuous Conversation Mode
+### Continuous Conversation Mode

 `conversation_mode` has ONLY two valid states:
- `"continuous"` — recommended for interactive agents
- Omit entirely — isolated per-node conversations (each node starts fresh)
+- `"continuous"` -- recommended (context carries across node transitions)
+- Omit entirely -- isolated per-node conversations

-**INVALID values** (do NOT use): `"client_facing"`, `"interactive"`,
-`"adaptive"`, `"shared"`. These do not exist in the framework.
-
-When `conversation_mode="continuous"`:
- Same conversation thread carries across node transitions
- Layered system prompts: identity (agent-level) + narrative + focus (per-node)
- Transition markers inserted at boundaries
- Compaction happens opportunistically at phase transitions
+**INVALID values:** `"client_facing"`, `"interactive"`, `"shared"`.

 ## loop_config

 Only three valid keys:
-```python
-loop_config = {
-    "max_iterations": 100,          # Max LLM turns per node visit
-    "max_tool_calls_per_turn": 20,  # Max tool calls per LLM response
-    "max_context_tokens": 32000,    # Triggers conversation compaction
+```json
+{
+  "max_iterations": 100,
+  "max_tool_calls_per_turn": 20,
+  "max_context_tokens": 32000
 }
 ```
-**INVALID keys** (do NOT use): `"strategy"`, `"mode"`, `"timeout"`,
-`"temperature"`. These are silently ignored or cause errors.

 ## Data Tools (Spillover)

 For large data that exceeds context:
- `save_data(filename, data)` — Write to session data dir
- `load_data(filename, offset, limit)` — Read with pagination
- `list_data_files()` — List files
- `serve_file_to_user(filename, label)` — Clickable file:// URI
+- `save_data(filename, data)` -- write to session data dir
+- `load_data(filename, offset, limit)` -- read with pagination
+- `list_data_files()` -- list files
+- `serve_file_to_user(filename, label)` -- clickable file URI

-`data_dir` is auto-injected by framework — LLM never sees it.
+`data_dir` is auto-injected by framework.

 ## Fan-Out / Fan-In

-Multiple ON_SUCCESS edges from same source → parallel execution via asyncio.gather().
- Parallel nodes must have disjoint output_keys
- Only one branch may have client_facing nodes
- Fan-in node gets all outputs in the shared buffer
+Multiple `on_success` edges from same source = parallel execution.
+Parallel nodes must have disjoint output_keys.

 ## Judge System

 - **Implicit** (default): ACCEPTs when LLM finishes with no tool calls and all required outputs set
 - **SchemaJudge**: Validates against Pydantic model
- **Custom**: Implement `evaluate(context) -> JudgeVerdict`
-
-Judge is the SOLE acceptance mechanism — no ad-hoc framework gating.
-
-## Triggers (Timers, Webhooks)
-
-For agents that react to external events, create a `triggers.json` file
-in the agent's export directory:
-
-```json
-[
-  {
-    "id": "daily-check",
-    "name": "Daily Check",
-    "trigger_type": "timer",
-    "trigger_config": {"cron": "0 9 * * *"},
-    "task": "Run the daily check process"
-  }
-]
-```
-
-### Key Fields
- `trigger_type`: `"timer"` or `"webhook"`
- `trigger_config`: `{"cron": "0 9 * * *"}` or `{"interval_minutes": 20}`
- `task`: describes what the worker should do when the trigger fires
- Triggers can also be created/removed at runtime via `set_trigger` / `remove_trigger` queen tools

 ## Tool Discovery

-Do NOT rely on a static tool list — it will be outdated. Always call
-`list_agent_tools()` with NO arguments first to see ALL available tools.
-Only use `group=` or `output_schema=` as follow-up calls after seeing the
-full list.
+Always call `list_agent_tools()` first to see available tools.
+Do NOT rely on a static tool list.

 ```
-list_agent_tools()                            # ALWAYS call this first
-list_agent_tools(group="gmail", output_schema="full")  # then drill into a category
-list_agent_tools("exports/my_agent/mcp_servers.json")  # specific agent's tools
+list_agent_tools()                                      # full summary
+list_agent_tools(group="gmail", output_schema="full")   # drill into category
 ```

-After building, run `validate_agent_package("{name}")` to check everything at once.
-
-Common tool categories (verify via list_agent_tools):
- **Web**: search, scrape, PDF
- **Data**: save/load/append/list data files, serve to user
- **File**: view, write, replace, diff, list, grep
- **Communication**: email, gmail, slack, telegram
- **CRM**: hubspot, apollo, calcom
- **GitHub**: stargazers, user profiles, repos
- **Vision**: image analysis
- **Time**: current time
+After building, run `validate_agent_package("{name}")` to check everything.
@@ -1,158 +1,80 @@
-# GCU Browser Automation Guide
+# Browser Automation Guide

-## When to Use GCU Nodes
+## When to Use Browser Nodes

-Use `node_type="gcu"` when:
- The user's workflow requires **navigating real websites** (scraping, form-filling, social media interaction, testing web UIs)
- The task involves **dynamic/JS-rendered pages** that `web_scrape` cannot handle (SPAs, infinite scroll, login-gated content)
- The agent needs to **interact with a website** — clicking, typing, scrolling, selecting, uploading files
+Use browser nodes (with `tools: {policy: "all"}`) when:
+- The task requires interacting with web pages (clicking, typing, navigating)
+- No API is available for the target service
+- The user is already logged in to the target site

-Do NOT use GCU for:
- Static content that `web_scrape` handles fine
- API-accessible data (use the API directly)
- PDF/file processing
- Anything that doesn't require a browser UI
+## What Browser Nodes Are

-## What GCU Nodes Are
+- Regular `event_loop` nodes with browser tools from gcu-tools MCP server
+- Set `tools: {policy: "all"}` to give access to all browser tools
+- Wire into the graph with edges like any other node
+- No special node_type needed

- `node_type="gcu"` — a declarative enhancement over `event_loop`
- Framework auto-prepends browser best-practices system prompt
- Framework auto-includes all 31 browser tools from `gcu-tools` MCP server
- Same underlying `EventLoopNode` class — no new imports needed
- `tools=[]` is correct — tools are auto-populated at runtime
+## Available Browser Tools

-## GCU Architecture Pattern  
+All tools are prefixed with `browser_`:
+- `browser_start`, `browser_open`, `browser_navigate` — launch/navigate
+- `browser_click`, `browser_click_coordinate`, `browser_fill`, `browser_type`, `browser_type_focused` — interact
+- `browser_press` (with optional `modifiers=["ctrl"]` etc.) — keyboard shortcuts
+- `browser_snapshot` — compact accessibility-tree read (structured)
+<!-- vision-only -->
+- `browser_screenshot` — visual capture (annotated PNG)
+<!-- /vision-only -->
+- `browser_shadow_query`, `browser_get_rect` — locate elements (shadow-piercing via `>>>`)
+- `browser_scroll`, `browser_wait` — navigation helpers
+- `browser_evaluate` — run JavaScript
+- `browser_close`, `browser_close_finished` — tab cleanup

-GCU nodes are **subagents** — invoked via `delegate_to_sub_agent()`, not connected via edges.
+## Pick the right reading tool

- Primary nodes (`event_loop`, client-facing) orchestrate; GCU nodes do browser work
- Parent node declares `sub_agents=["gcu-node-id"]` and calls `delegate_to_sub_agent(agent_id="gcu-node-id", task="...")`
- GCU nodes set `max_node_visits=1` (single execution per delegation), `client_facing=False`
- GCU nodes use `output_keys=["result"]` and return structured JSON via `set_output("result", ...)`
+**`browser_snapshot`** — compact accessibility tree of interactive elements. Fast, cheap, good for static or form-heavy pages where the DOM matches what's visually rendered (documentation, simple dashboards, search results, settings pages).

-## GCU Node Definition Template
+**`browser_screenshot`** — visual capture + metadata (`cssWidth`, `devicePixelRatio`, scale fields). Use this when `browser_snapshot` does not show the thing you need, when refs look stale, or when visual position/layout matters. This often happens on complex SPAs — LinkedIn, Twitter/X, Reddit, Gmail, Notion, Slack, Discord — and on sites using shadow DOM, virtual scrolling, React reconciliation, or dynamic layout.

-```python
-gcu_browser_node = NodeSpec(
-    id="gcu-browser-worker",
-    name="Browser Worker",
-    description="Browser subagent that does X.",
-    node_type="gcu",
-    client_facing=False,
-    max_node_visits=1,
-    input_keys=[],
-    output_keys=["result"],
-    tools=[],  # Auto-populated with all browser tools
-    system_prompt="""\
-You are a browser agent. Your job: [specific task].
+Neither tool is "preferred" universally — they're for different jobs. Start with snapshot for page structure and ordinary controls; use screenshot as the fallback when snapshot can't find or verify the visible target. Activate the `browser-automation` skill for the full decision tree.

-## Workflow
-1. browser_start (only if no browser is running yet)
-2. browser_open(url=TARGET_URL) — note the returned targetId
-3. browser_snapshot to read the page
-4. [task-specific steps]
-5. set_output("result", JSON)
+## Coordinate rule

-## Output format
-set_output("result", JSON) with:
- [field]: [type and description]
-""",
-)
+Every browser tool that takes or returns coordinates operates in **fractions of the viewport (0..1 for both axes)**. Read a target's proportional position off `browser_screenshot` ("~35% from the left, ~20% from the top" → `(0.35, 0.20)`) and pass that to `browser_click_coordinate` / `browser_hover_coordinate` / `browser_press_at`. `browser_get_rect` and `browser_shadow_query` return `rect.cx` / `rect.cy` as fractions. The tools multiply by `cssWidth` / `cssHeight` internally — no scale awareness required. Fractions are used because every vision model (Claude, GPT-4o, Gemini, local VLMs) resizes/tiles images differently; proportions are invariant. Avoid raw `getBoundingClientRect()` via `browser_evaluate` for coord lookup; use `browser_get_rect` instead.
+
+## System prompt tips for browser nodes
+
+```
+1. Start with browser_snapshot or the snapshot returned by the latest interaction.
+2. If the target is missing, ambiguous, stale, or visibly present but absent from the tree,
+   use browser_screenshot to orient and then click by fractional coordinates.
+3. Before typing into a rich-text editor (X compose, LinkedIn DM, Gmail, Reddit),
+   click the input area first with browser_click_coordinate so React / Draft.js /
+   Lexical register a native focus event, then use browser_type_focused(text=...)
+   for shadow-DOM inputs or browser_type(selector, text) for light-DOM inputs.
+4. Use browser_wait(seconds=2-3) after navigation for SPA hydration.
+5. If you hit an auth wall, call set_output with an error and move on.
+6. Keep tool calls per turn <= 10 for reliability.
 ```

-## Parent Node Template (orchestrating GCU subagents)
-
-```python
-orchestrator_node = NodeSpec(
-    id="orchestrator",
-    ...
-    node_type="event_loop",
-    sub_agents=["gcu-browser-worker"],
-    system_prompt="""\
-...
-delegate_to_sub_agent(
-    agent_id="gcu-browser-worker",
-    task="Navigate to [URL]. Do [specific task]. Return JSON with [fields]."
-)
-...
-""",
-    tools=[],  # Orchestrator doesn't need browser tools
-)
-```
-
-## mcp_servers.json with GCU
+## Example

 ```json
 {
-  "hive-tools": { ... },
-  "gcu-tools": {
-    "transport": "stdio",
-    "command": "uv",
-    "args": ["run", "python", "-m", "gcu.server", "--stdio"],
-    "cwd": "../../tools",
-    "description": "GCU tools for browser automation"
-  }
+  "id": "scan-profiles",
+  "name": "Scan LinkedIn Profiles",
+  "description": "Navigate LinkedIn search results and collect profile data",
+  "tools": {"policy": "all"},
+  "input_keys": ["search_url"],
+  "output_keys": ["profiles"],
+  "system_prompt": "Navigate to the search URL via browser_navigate(wait_until='load', timeout_ms=20000). Wait 3s for SPA hydration. Use the returned snapshot to look for result cards first. If the cards are missing, stale, or visually present but absent from the tree, use browser_screenshot to orient; paginate through results by scrolling and use screenshots only when the snapshot cannot find or verify the visible cards..."
 }
 ```

-Note: `gcu-tools` is auto-added if any node uses `node_type="gcu"`, but including it explicitly is fine.
-
-## GCU System Prompt Best Practices
-
-Key rules to bake into GCU node prompts:
-
- Prefer `browser_snapshot` over `browser_get_text("body")` — compact accessibility tree vs 100KB+ raw HTML
- Always `browser_wait` after navigation
- Use large scroll amounts (~2000-5000) for lazy-loaded content
- For spillover files, use `run_command` with grep, not `read_file`
- If auth wall detected, report immediately — don't attempt login
- Keep tool calls per turn ≤10
- Tab isolation: when browser is already running, use `browser_open(background=true)` and pass `target_id` to every call
-
-## Multiple Concurrent GCU Subagents
-
-When a task can be parallelized across multiple sites or profiles, declare a distinct GCU
-node for each and invoke them all in the same LLM turn.  The framework batches all
-`delegate_to_sub_agent` calls made in one turn and runs them with `asyncio.gather`, so
-they execute concurrently — not sequentially.
-
-**Each GCU subagent automatically gets its own isolated browser context** — no `profile=`
-argument is needed in tool calls.  The framework derives a unique profile from the subagent's
-node ID and instance counter and injects it via an asyncio `ContextVar` before the subagent
-runs.
-
-### Example: three sites in parallel
-
-```python
-# Three distinct GCU nodes
-gcu_site_a = NodeSpec(id="gcu-site-a", node_type="gcu", ...)
-gcu_site_b = NodeSpec(id="gcu-site-b", node_type="gcu", ...)
-gcu_site_c = NodeSpec(id="gcu-site-c", node_type="gcu", ...)
-
-orchestrator = NodeSpec(
-    id="orchestrator",
-    node_type="event_loop",
-    sub_agents=["gcu-site-a", "gcu-site-b", "gcu-site-c"],
-    system_prompt="""\
-Call all three subagents in a single response to run them in parallel:
-  delegate_to_sub_agent(agent_id="gcu-site-a", task="Scrape prices from site A")
-  delegate_to_sub_agent(agent_id="gcu-site-b", task="Scrape prices from site B")
-  delegate_to_sub_agent(agent_id="gcu-site-c", task="Scrape prices from site C")
-""",
-)
+Connected via regular edges:
+```
+search-setup -> scan-profiles -> process-results
 ```

-**Rules:**
- Use distinct node IDs for each concurrent task — sharing an ID shares the browser context.
- The GCU node prompts do not need to mention `profile=`; isolation is automatic.
- Cleanup is automatic at session end, but GCU nodes can call `browser_stop()` explicitly
-  if they want to release resources mid-run.
+## Further detail

-## GCU Anti-Patterns
-
- Using `browser_screenshot` to read text (use `browser_snapshot` instead; screenshots are for visual context only)
- Re-navigating after scrolling (resets scroll position)
- Attempting login on auth walls
- Forgetting `target_id` in multi-tab scenarios
- Putting browser tools directly on `event_loop` nodes instead of using GCU subagent pattern
- Making GCU nodes `client_facing=True` (they should be autonomous subagents)
+For rich-text editor quirks (Lexical, Draft.js, ProseMirror), shadow-DOM shortcuts, `beforeunload` dialog neutralization, Trusted Types CSP on LinkedIn, keyboard shortcut dispatch, and per-site selector tables — **activate the `browser-automation` skill**. That skill has the full verified guidance and is refreshed against real production sites.
@@ -22,10 +22,10 @@ def mock_mode():

@pytest_asyncio.fixture(scope="session")
 async def runner(tmp_path_factory, mock_mode):
-    from framework.runner.runner import AgentRunner
+    from framework.loader.agent_loader import AgentLoader

    storage = tmp_path_factory.mktemp("agent_storage")
-    r = AgentRunner.load(AGENT_PATH, mock_mode=mock_mode, storage_path=storage)
+    r = AgentLoader.load(AGENT_PATH, mock_mode=mock_mode, storage_path=storage)
    r._setup()
    yield r
    await r.cleanup_async()
@@ -2,17 +2,22 @@
 Command-line interface for Aden Hive.

 Usage:
-    hive run exports/my-agent --input '{"key": "value"}'
-    hive info exports/my-agent
-    hive validate exports/my-agent
-    hive list exports/
-    hive shell exports/my-agent
+    hive serve                       Start the HTTP API server
+    hive open                        Start the server and open the dashboard
+    hive queen list                  List queen profiles
+    hive queen show <queen_id>       Inspect a queen profile
+    hive queen sessions <queen_id>   List a queen's sessions
+    hive colony list                 List colonies on disk
+    hive colony info <name>          Inspect a colony
+    hive colony delete <name>        Delete a colony
+    hive session list                List live sessions (use --cold for on-disk)
+    hive session stop <session_id>   Stop a live session
+    hive chat <session_id> "msg"     Send a message to a live queen

-Testing commands:
-    hive test-run <agent_path> --goal <goal_id>
-    hive test-debug <agent_path> <test_name>
-    hive test-list <agent_path>
-    hive test-stats <agent_path>
+Subsystems:
+    hive skill ...                   Manage skills (~/.hive/skills/)
+    hive mcp ...                     Manage MCP servers
+    hive debugger                    LLM debug log viewer
 """

 import argparse
@@ -20,86 +25,57 @@ import sys
 from pathlib import Path


-def _configure_paths():
-    """Auto-configure sys.path so agents in exports/ are discoverable.
+def _configure_paths() -> None:
+    """Auto-configure sys.path so the framework is importable from any cwd.

-    Resolves the project root by walking up from this file (framework/cli.py lives
-    inside core/framework/) or from CWD, then adds the exports/ directory to sys.path
-    if it exists. This eliminates the need for manual PYTHONPATH configuration.
+    Walks up from this file to find the project root, then ensures
+    `core/` is on sys.path so `framework.*` imports resolve when the
+    package isn't installed via `pip install -e .`.
    """
-    # Strategy 1: resolve relative to this file (works when installed via pip install -e core/)
    framework_dir = Path(__file__).resolve().parent  # core/framework/
    core_dir = framework_dir.parent  # core/
    project_root = core_dir.parent  # project root

-    # Strategy 2: if project_root doesn't look right, fall back to CWD
-    if not (project_root / "exports").is_dir() and not (project_root / "core").is_dir():
+    if not (project_root / "core").is_dir():
        project_root = Path.cwd()

-    # Add exports/ to sys.path so agents are importable as top-level packages
-    exports_dir = project_root / "exports"
-    if exports_dir.is_dir():
-        exports_str = str(exports_dir)
-        if exports_str not in sys.path:
-            sys.path.insert(0, exports_str)
-
-    # Add examples/templates/ to sys.path so template agents are importable
-    templates_dir = project_root / "examples" / "templates"
-    if templates_dir.is_dir():
-        templates_str = str(templates_dir)
-        if templates_str not in sys.path:
-            sys.path.insert(0, templates_str)
-
-    # Ensure core/ is also in sys.path (for non-editable-install scenarios)
    core_str = str(project_root / "core")
    if (project_root / "core").is_dir() and core_str not in sys.path:
        sys.path.insert(0, core_str)

-    # Add core/framework/agents/ so framework agents are importable as top-level packages
-    framework_agents_dir = project_root / "core" / "framework" / "agents"
-    if framework_agents_dir.is_dir():
-        fa_str = str(framework_agents_dir)
-        if fa_str not in sys.path:
-            sys.path.insert(0, fa_str)

-
-def main():
+def main() -> None:
    _configure_paths()

    parser = argparse.ArgumentParser(
        prog="hive",
-        description="Aden Hive - Build and run goal-driven agents",
+        description="Aden Hive — Queens, colonies, and live agent sessions",
    )
    parser.add_argument(
        "--model",
        default="claude-haiku-4-5-20251001",
-        help="Anthropic model to use",
+        help="Default LLM model (Anthropic ID)",
    )

    subparsers = parser.add_subparsers(dest="command", required=True)

-    # Register runner commands (run, info, validate, list, shell)
-    from framework.runner.cli import register_commands
+    # Core commands: serve, open, queen, colony, session, chat
+    from framework.loader.cli import register_commands

    register_commands(subparsers)

-    # Register testing commands (test-run, test-debug, test-list, test-stats)
-    from framework.testing.cli import register_testing_commands
-
-    register_testing_commands(subparsers)
-
-    # Register skill commands (skill list, skill trust, ...)
+    # Skill management (~/.hive/skills/)
    from framework.skills.cli import register_skill_commands

    register_skill_commands(subparsers)

-    # Register debugger commands (debugger)
+    # LLM debug log viewer
    from framework.debugger.cli import register_debugger_commands

    register_debugger_commands(subparsers)

-    # Register MCP registry commands (mcp install, mcp add, ...)
-    from framework.runner.mcp_registry_cli import register_mcp_commands
+    # MCP server registry
+    from framework.loader.mcp_registry_cli import register_mcp_commands

    register_mcp_commands(subparsers)

@@ -12,13 +12,47 @@ from dataclasses import dataclass, field
 from pathlib import Path
 from typing import Any

-from framework.graph.edge import DEFAULT_MAX_TOKENS
+DEFAULT_MAX_TOKENS = 8192
+
+# ---------------------------------------------------------------------------
+# Hive home directory structure
+# ---------------------------------------------------------------------------
+
+HIVE_HOME = Path.home() / ".hive"
+QUEENS_DIR = HIVE_HOME / "agents" / "queens"
+COLONIES_DIR = HIVE_HOME / "colonies"
+MEMORIES_DIR = HIVE_HOME / "memories"
+
+
+def queen_dir(queen_name: str = "default") -> Path:
+    """Return the storage directory for a named queen agent."""
+    return QUEENS_DIR / queen_name
+
+
+def colony_dir(colony_name: str) -> Path:
+    """Return the directory for a named colony."""
+    return COLONIES_DIR / colony_name
+
+
+def memory_dir(scope: str, name: str | None = None) -> Path:
+    """Return memory dir for a scope.
+
+    Examples::
+
+        memory_dir("global")                  -> ~/.hive/memories/global
+        memory_dir("colonies", "my_agent")    -> ~/.hive/memories/colonies/my_agent
+        memory_dir("agents/queens", "default")-> ~/.hive/memories/agents/queens/default
+        memory_dir("agents", "worker_name")   -> ~/.hive/memories/agents/worker_name
+    """
+    base = MEMORIES_DIR / scope
+    return base / name if name else base
+

 # ---------------------------------------------------------------------------
 # Low-level config file access
 # ---------------------------------------------------------------------------

-HIVE_CONFIG_FILE = Path.home() / ".hive" / "configuration.json"
+HIVE_CONFIG_FILE = HIVE_HOME / "configuration.json"

 # Hive LLM router endpoint (Anthropic-compatible).
 # litellm's Anthropic handler appends /v1/messages, so this is just the base host.
@@ -42,6 +76,48 @@ def get_hive_config() -> dict[str, Any]:
        return {}


+# ---------------------------------------------------------------------------
+# Credential store helpers (for BYOK keys)
+# ---------------------------------------------------------------------------
+
+# Provider name → credential store ID mapping
+_PROVIDER_CRED_MAP: dict[str, str] = {
+    "anthropic": "anthropic",
+    "openai": "openai",
+    "gemini": "gemini",
+    "google": "gemini",
+    "minimax": "minimax",
+    "groq": "groq",
+    "cerebras": "cerebras",
+    "openrouter": "openrouter",
+    "mistral": "mistral",
+    "together": "together",
+    "together_ai": "together",
+    "deepseek": "deepseek",
+    "kimi": "kimi",
+    "hive": "hive",
+}
+
+
+def _get_api_key_from_credential_store(provider: str) -> str | None:
+    """Look up a BYOK API key from the encrypted credential store.
+
+    Returns None if no key is found or the credential store is unavailable.
+    """
+    if not os.environ.get("HIVE_CREDENTIAL_KEY"):
+        return None
+    cred_id = _PROVIDER_CRED_MAP.get(provider.lower())
+    if not cred_id:
+        return None
+    try:
+        from framework.credentials import CredentialStore
+
+        store = CredentialStore.with_encrypted_storage()
+        return store.get(cred_id)
+    except Exception:
+        return None
+
+
 # ---------------------------------------------------------------------------
 # Derived helpers
 # ---------------------------------------------------------------------------
@@ -88,7 +164,7 @@ def get_worker_api_key() -> str | None:
    # Worker-specific subscription / env var
    if worker_llm.get("use_claude_code_subscription"):
        try:
-            from framework.runner.runner import get_claude_code_token
+            from framework.loader.agent_loader import get_claude_code_token

            token = get_claude_code_token()
            if token:
@@ -98,7 +174,7 @@ def get_worker_api_key() -> str | None:

    if worker_llm.get("use_codex_subscription"):
        try:
-            from framework.runner.runner import get_codex_token
+            from framework.loader.agent_loader import get_codex_token

            token = get_codex_token()
            if token:
@@ -108,7 +184,7 @@ def get_worker_api_key() -> str | None:

    if worker_llm.get("use_kimi_code_subscription"):
        try:
-            from framework.runner.runner import get_kimi_code_token
+            from framework.loader.agent_loader import get_kimi_code_token

            token = get_kimi_code_token()
            if token:
@@ -118,7 +194,7 @@ def get_worker_api_key() -> str | None:

    if worker_llm.get("use_antigravity_subscription"):
        try:
-            from framework.runner.runner import get_antigravity_token
+            from framework.loader.agent_loader import get_antigravity_token

            token = get_antigravity_token()
            if token:
@@ -174,7 +250,7 @@ def get_worker_llm_extra_kwargs() -> dict[str, Any]:
                "User-Agent": "CodexBar",
            }
            try:
-                from framework.runner.runner import get_codex_account_id
+                from framework.loader.agent_loader import get_codex_account_id

                account_id = get_codex_account_id()
                if account_id:
@@ -221,22 +297,43 @@ def get_max_context_tokens() -> int:
    return get_hive_config().get("llm", {}).get("max_context_tokens", DEFAULT_MAX_CONTEXT_TOKENS)


+def get_api_keys() -> list[str] | None:
+    """Return a list of API keys if ``api_keys`` is configured, else ``None``.
+
+    This supports key-pool rotation: configure multiple keys in
+    ``~/.hive/configuration.json`` under ``llm.api_keys`` and the
+    :class:`~framework.llm.key_pool.KeyPool` will rotate through them.
+    """
+    llm = get_hive_config().get("llm", {})
+    keys = llm.get("api_keys")
+    if keys and isinstance(keys, list) and len(keys) > 0:
+        return [k for k in keys if k]  # filter empties
+    return None
+
+
 def get_api_key() -> str | None:
    """Return the API key, supporting env var, Claude Code subscription, Codex, and ZAI Code.

    Priority:
+    0. Explicit key pool (``api_keys`` list) -- returns first key for
+       single-key callers; full pool available via :func:`get_api_keys`.
    1. Claude Code subscription (``use_claude_code_subscription: true``)
       reads the OAuth token from ``~/.claude/.credentials.json``.
    2. Codex subscription (``use_codex_subscription: true``)
       reads the OAuth token from macOS Keychain or ``~/.codex/auth.json``.
    3. Environment variable named in ``api_key_env_var``.
    """
+    # If an explicit key pool is configured, use the first key.
+    pool_keys = get_api_keys()
+    if pool_keys:
+        return pool_keys[0]
+
    llm = get_hive_config().get("llm", {})

    # Claude Code subscription: read OAuth token directly
    if llm.get("use_claude_code_subscription"):
        try:
-            from framework.runner.runner import get_claude_code_token
+            from framework.loader.agent_loader import get_claude_code_token

            token = get_claude_code_token()
            if token:
@@ -247,7 +344,7 @@ def get_api_key() -> str | None:
    # Codex subscription: read OAuth token from Keychain / auth.json
    if llm.get("use_codex_subscription"):
        try:
-            from framework.runner.runner import get_codex_token
+            from framework.loader.agent_loader import get_codex_token

            token = get_codex_token()
            if token:
@@ -258,7 +355,7 @@ def get_api_key() -> str | None:
    # Kimi Code subscription: read API key from ~/.kimi/config.toml
    if llm.get("use_kimi_code_subscription"):
        try:
-            from framework.runner.runner import get_kimi_code_token
+            from framework.loader.agent_loader import get_kimi_code_token

            token = get_kimi_code_token()
            if token:
@@ -269,7 +366,7 @@ def get_api_key() -> str | None:
    # Antigravity subscription: read OAuth token from accounts JSON
    if llm.get("use_antigravity_subscription"):
        try:
-            from framework.runner.runner import get_antigravity_token
+            from framework.loader.agent_loader import get_antigravity_token

            token = get_antigravity_token()
            if token:
@@ -280,8 +377,12 @@ def get_api_key() -> str | None:
    # Standard env-var path (covers ZAI Code and all API-key providers)
    api_key_env_var = llm.get("api_key_env_var")
    if api_key_env_var:
-        return os.environ.get(api_key_env_var)
-    return None
+        key = os.environ.get(api_key_env_var)
+        if key:
+            return key
+
+    # Credential store fallback — BYOK keys stored via the UI
+    return _get_api_key_from_credential_store(llm.get("provider", ""))


 # OAuth credentials for Antigravity are fetched from the opencode-antigravity-auth project.
@@ -304,9 +405,7 @@ def _fetch_antigravity_credentials() -> tuple[str | None, str | None]:
    import urllib.request

    try:
-        req = urllib.request.Request(
-            _ANTIGRAVITY_CREDENTIALS_URL, headers={"User-Agent": "Hive/1.0"}
-        )
+        req = urllib.request.Request(_ANTIGRAVITY_CREDENTIALS_URL, headers={"User-Agent": "Hive/1.0"})
        with urllib.request.urlopen(req, timeout=10) as resp:
            content = resp.read().decode("utf-8")
            id_match = re.search(r'ANTIGRAVITY_CLIENT_ID\s*=\s*"([^"]+)"', content)
@@ -422,7 +521,7 @@ def get_llm_extra_kwargs() -> dict[str, Any]:
                "User-Agent": "CodexBar",
            }
            try:
-                from framework.runner.runner import get_codex_account_id
+                from framework.loader.agent_loader import get_codex_account_id

                account_id = get_codex_account_id()
                if account_id:
@@ -51,6 +51,7 @@ from .key_storage import (
 from .models import (
    CredentialDecryptionError,
    CredentialError,
+    CredentialExpiredError,
    CredentialKey,
    CredentialKeyNotFoundError,
    CredentialNotFoundError,
@@ -84,6 +85,7 @@ from .template import TemplateResolver
 from .validation import (
    CredentialStatus,
    CredentialValidationResult,
+    compute_unavailable_tools,
    ensure_credential_key_env,
    validate_agent_credentials,
 )
@@ -136,6 +138,7 @@ __all__ = [
    "CredentialNotFoundError",
    "CredentialKeyNotFoundError",
    "CredentialRefreshError",
+    "CredentialExpiredError",
    "CredentialValidationError",
    "CredentialDecryptionError",
    # Key storage (bootstrap credentials)
@@ -148,6 +151,7 @@ __all__ = [
    # Validation
    "ensure_credential_key_env",
    "validate_agent_credentials",
+    "compute_unavailable_tools",
    "CredentialStatus",
    "CredentialValidationResult",
    # Interactive setup
@@ -332,9 +332,7 @@ class AdenCredentialClient:
                last_error = e
                if attempt < self.config.retry_attempts - 1:
                    delay = self.config.retry_delay * (2**attempt)
-                    logger.warning(
-                        f"Aden request failed (attempt {attempt + 1}), retrying in {delay}s: {e}"
-                    )
+                    logger.warning(f"Aden request failed (attempt {attempt + 1}), retrying in {delay}s: {e}")
                    time.sleep(delay)
                else:
                    raise AdenClientError(f"Failed to connect to Aden server: {e}") from e
@@ -347,9 +345,7 @@ class AdenCredentialClient:
            ):
                raise

-        raise AdenClientError(
-            f"Request failed after {self.config.retry_attempts} attempts"
-        ) from last_error
+        raise AdenClientError(f"Request failed after {self.config.retry_attempts} attempts") from last_error

    def list_integrations(self) -> list[AdenIntegrationInfo]:
        """
@@ -192,9 +192,7 @@ class AdenSyncProvider(CredentialProvider):
                    f"Visit: {e.reauthorization_url or 'your Aden dashboard'}"
                ) from e

-            raise CredentialRefreshError(
-                f"Failed to refresh credential '{credential.id}': {e}"
-            ) from e
+            raise CredentialRefreshError(f"Failed to refresh credential '{credential.id}': {e}") from e

        except AdenClientError as e:
            logger.error(f"Aden client error for '{credential.id}': {e}")
@@ -206,9 +204,7 @@ class AdenSyncProvider(CredentialProvider):
                    logger.warning(f"Aden unavailable, using cached token for '{credential.id}'")
                    return credential

-            raise CredentialRefreshError(
-                f"Aden server unavailable and token expired for '{credential.id}'"
-            ) from e
+            raise CredentialRefreshError(f"Aden server unavailable and token expired for '{credential.id}'") from e

    def validate(self, credential: CredentialObject) -> bool:
        """
@@ -168,9 +168,7 @@ class AdenCachedStorage(CredentialStorage):
                if rid != credential_id:
                    result = self._load_by_id(rid)
                    if result is not None:
-                        logger.info(
-                            f"Loaded credential '{credential_id}' via provider index (id='{rid}')"
-                        )
+                        logger.info(f"Loaded credential '{credential_id}' via provider index (id='{rid}')")
                        return result

        # Direct lookup (exact credential_id match)
@@ -199,6 +197,19 @@ class AdenCachedStorage(CredentialStorage):
        if local_cred is None:
            return None

+        # Skip Aden fetch for credentials not managed by Aden (BYOK credentials).
+        # Only OAuth credentials synced from Aden are in the provider index.
+        # BYOK credentials like anthropic, brave_search are local-only.
+        # Also check the _aden_managed flag on the credential itself.
+        is_aden_managed = (
+            credential_id in self._provider_index
+            or any(credential_id in ids for ids in self._provider_index.values())
+            or (local_cred is not None and local_cred.keys.get("_aden_managed") is not None)
+        )
+        if not is_aden_managed:
+            logger.debug(f"Credential '{credential_id}' is local-only, skipping Aden refresh")
+            return local_cred
+
        # Try to refresh stale local credential from Aden
        try:
            aden_cred = self._aden_provider.fetch_from_aden(credential_id)
@@ -493,9 +493,7 @@ class TestAdenCachedStorage:
        assert loaded is not None
        assert loaded.keys["access_token"].value.get_secret_value() == "cached-token"

-    def test_load_from_aden_when_stale(
-        self, cached_storage, local_storage, provider, mock_client, aden_response
-    ):
+    def test_load_from_aden_when_stale(self, cached_storage, local_storage, provider, mock_client, aden_response):
        """Test load fetches from Aden when cache is stale."""
        # Create stale cached credential
        cred = CredentialObject(
@@ -521,9 +519,7 @@ class TestAdenCachedStorage:
        assert loaded is not None
        assert loaded.keys["access_token"].value.get_secret_value() == "test-access-token"

-    def test_load_falls_back_to_stale_when_aden_fails(
-        self, cached_storage, local_storage, provider, mock_client
-    ):
+    def test_load_falls_back_to_stale_when_aden_fails(self, cached_storage, local_storage, provider, mock_client):
        """Test load falls back to stale cache when Aden fails."""
        # Create stale cached credential
        cred = CredentialObject(
@@ -333,6 +333,29 @@ class CredentialRefreshError(CredentialError):
    pass


+class CredentialExpiredError(CredentialError):
+    """Raised when a credential is expired and refresh has failed.
+
+    Carries the metadata an agent (or the tool runner) needs to surface a
+    reauth request to the user without having to look anything else up.
+    """
+
+    def __init__(
+        self,
+        credential_id: str,
+        message: str,
+        *,
+        provider: str | None = None,
+        alias: str | None = None,
+        help_url: str | None = None,
+    ):
+        self.credential_id = credential_id
+        self.provider = provider
+        self.alias = alias
+        self.help_url = help_url
+        super().__init__(message)
+
+
 class CredentialValidationError(CredentialError):
    """Raised when credential validation fails."""

@@ -95,9 +95,7 @@ class BaseOAuth2Provider(CredentialProvider):

                self._client = httpx.Client(timeout=self.config.request_timeout)
            except ImportError as e:
-                raise ImportError(
-                    "OAuth2 provider requires 'httpx'. Install with: uv pip install httpx"
-                ) from e
+                raise ImportError("OAuth2 provider requires 'httpx'. Install with: uv pip install httpx") from e
        return self._client

    def _close_client(self) -> None:
@@ -311,8 +309,7 @@ class BaseOAuth2Provider(CredentialProvider):
        except OAuth2Error as e:
            if e.error == "invalid_grant":
                raise CredentialRefreshError(
-                    f"Refresh token for '{credential.id}' is invalid or revoked. "
-                    "Re-authorization required."
+                    f"Refresh token for '{credential.id}' is invalid or revoked. Re-authorization required."
                ) from e
            raise CredentialRefreshError(f"Failed to refresh '{credential.id}': {e}") from e

@@ -422,9 +419,7 @@ class BaseOAuth2Provider(CredentialProvider):
        if response.status_code != 200 or "error" in response_data:
            error = response_data.get("error", "unknown_error")
            description = response_data.get("error_description", response.text)
-            raise OAuth2Error(
-                error=error, description=description, status_code=response.status_code
-            )
+            raise OAuth2Error(error=error, description=description, status_code=response.status_code)

        return OAuth2Token.from_token_response(response_data)

@@ -158,9 +158,7 @@ class TokenLifecycleManager:
        """
        # Run in executor to avoid blocking
        loop = asyncio.get_event_loop()
-        token = await loop.run_in_executor(
-            None, lambda: self.provider.client_credentials_grant(scopes=scopes)
-        )
+        token = await loop.run_in_executor(None, lambda: self.provider.client_credentials_grant(scopes=scopes))

        self._save_token_to_store(token)
        self._cached_token = token
@@ -100,9 +100,7 @@ class ZohoOAuth2Provider(BaseOAuth2Provider):
        )
        super().__init__(config, provider_id="zoho_crm_oauth2")
        self._accounts_domain = base
-        self._api_domain = (
-            api_domain or os.getenv("ZOHO_API_DOMAIN", "https://www.zohoapis.com")
-        ).rstrip("/")
+        self._api_domain = (api_domain or os.getenv("ZOHO_API_DOMAIN", "https://www.zohoapis.com")).rstrip("/")

    @property
    def supported_types(self) -> list[CredentialType]:
@@ -36,7 +36,7 @@ from pathlib import Path
 from typing import TYPE_CHECKING, Any

 if TYPE_CHECKING:
-    from framework.graph import NodeSpec
+    from framework.orchestrator import NodeSpec

 logger = logging.getLogger(__name__)

@@ -268,9 +268,7 @@ class CredentialSetupSession:
        self._print(f"{Colors.YELLOW}Initializing credential store...{Colors.NC}")
        try:
            generate_and_save_credential_key()
-            self._print(
-                f"{Colors.GREEN}✓ Encryption key saved to ~/.hive/secrets/credential_key{Colors.NC}"
-            )
+            self._print(f"{Colors.GREEN}✓ Encryption key saved to ~/.hive/secrets/credential_key{Colors.NC}")
            return True
        except Exception as e:
            self._print(f"{Colors.RED}Failed to initialize credential store: {e}{Colors.NC}")
@@ -449,9 +447,7 @@ class CredentialSetupSession:
                    logger.warning("Unexpected error exporting credential to env", exc_info=True)
                return True
            else:
-                self._print(
-                    f"{Colors.YELLOW}⚠ {cred.credential_name} not found in Aden account.{Colors.NC}"
-                )
+                self._print(f"{Colors.YELLOW}⚠ {cred.credential_name} not found in Aden account.{Colors.NC}")
                self._print("Please connect this integration on https://hive.adenhq.com first.")
                return False
        except Exception as e:
@@ -533,7 +529,9 @@ class CredentialSetupSession:


 def load_agent_nodes(agent_path: str | Path) -> list:
-    """Load NodeSpec list from an agent's agent.py or agent.json.
+    """Load NodeSpec list from an agent directory.
+
+    Checks agent.json (declarative) first, then agent.py (legacy).

    Args:
        agent_path: Path to agent directory.
@@ -542,16 +540,28 @@ def load_agent_nodes(agent_path: str | Path) -> list:
        List of NodeSpec objects (empty list if agent can't be loaded).
    """
    agent_path = Path(agent_path)
+    agent_json_file = agent_path / "agent.json"
    agent_py = agent_path / "agent.py"
-    agent_json = agent_path / "agent.json"

-    if agent_py.exists():
+    if agent_json_file.exists():
+        return _load_nodes_from_json_declarative(agent_json_file)
+    elif agent_py.exists():
        return _load_nodes_from_python_agent(agent_path)
-    elif agent_json.exists():
-        return _load_nodes_from_json_agent(agent_json)
    return []


+def _load_nodes_from_json_declarative(agent_json: Path) -> list:
+    """Load nodes from a declarative JSON agent."""
+    try:
+        from framework.loader.agent_loader import load_agent_config
+
+        data = json.loads(agent_json.read_text(encoding="utf-8"))
+        graph, _ = load_agent_config(data)
+        return list(graph.nodes)
+    except Exception:
+        return []
+
+
 def _load_nodes_from_python_agent(agent_path: Path) -> list:
    """Load nodes from a Python-based agent."""
    import importlib.util
@@ -590,7 +600,7 @@ def _load_nodes_from_json_agent(agent_json: Path) -> list:
        with open(agent_json, encoding="utf-8-sig") as f:
            data = json.load(f)

-        from framework.graph import NodeSpec
+        from framework.orchestrator import NodeSpec

        nodes_data = data.get("graph", {}).get("nodes", [])
        nodes = []
@@ -136,8 +136,7 @@ class EncryptedFileStorage(CredentialStorage):
            from cryptography.fernet import Fernet
        except ImportError as e:
            raise ImportError(
-                "Encrypted storage requires 'cryptography'. "
-                "Install with: uv pip install cryptography"
+                "Encrypted storage requires 'cryptography'. Install with: uv pip install cryptography"
            ) from e

        self.base_path = Path(base_path or self.DEFAULT_PATH).expanduser()
@@ -161,6 +160,14 @@ class EncryptedFileStorage(CredentialStorage):

        self._fernet = Fernet(self._key)

+        # Rebuild the metadata index from disk if it's missing or older than
+        # the current index schema. The index is a developer-readable JSON
+        # snapshot of the encrypted store; the .enc files remain authoritative.
+        try:
+            self._maybe_rebuild_index()
+        except Exception:
+            logger.debug("Initial index rebuild failed (non-fatal)", exc_info=True)
+
    def _ensure_dirs(self) -> None:
        """Create directory structure."""
        (self.base_path / "credentials").mkdir(parents=True, exist_ok=True)
@@ -186,8 +193,8 @@ class EncryptedFileStorage(CredentialStorage):
        with open(cred_path, "wb") as f:
            f.write(encrypted)

-        # Update index
-        self._update_index(credential.id, "save", credential.credential_type.value)
+        # Update developer-readable index
+        self._index_upsert(credential)
        logger.debug(f"Saved encrypted credential '{credential.id}'")

    def load(self, credential_id: str) -> CredentialObject | None:
@@ -205,9 +212,7 @@ class EncryptedFileStorage(CredentialStorage):
            json_bytes = self._fernet.decrypt(encrypted)
            data = json.loads(json_bytes.decode("utf-8-sig"))
        except Exception as e:
-            raise CredentialDecryptionError(
-                f"Failed to decrypt credential '{credential_id}': {e}"
-            ) from e
+            raise CredentialDecryptionError(f"Failed to decrypt credential '{credential_id}': {e}") from e

        # Deserialize
        return self._deserialize_credential(data)
@@ -217,7 +222,7 @@ class EncryptedFileStorage(CredentialStorage):
        cred_path = self._cred_path(credential_id)
        if cred_path.exists():
            cred_path.unlink()
-            self._update_index(credential_id, "delete")
+            self._index_remove(credential_id)
            logger.debug(f"Deleted credential '{credential_id}'")
            return True
        return False
@@ -258,33 +263,151 @@ class EncryptedFileStorage(CredentialStorage):

        return CredentialObject.model_validate(data)

-    def _update_index(
-        self,
-        credential_id: str,
-        operation: str,
-        credential_type: str | None = None,
-    ) -> None:
-        """Update the metadata index."""
-        index_path = self.base_path / "metadata" / "index.json"
+    # ------------------------------------------------------------------
+    # Developer-readable metadata index
+    #
+    # The index lives at ``<base_path>/metadata/index.json`` and mirrors what
+    # is in the encrypted store at a glance: credential id, provider, alias,
+    # identity, key names, timestamps, and earliest expiry. It contains NO
+    # secret values and is safe to share when filing a bug report. The .enc
+    # files remain authoritative — the index is purely for human inspection
+    # and for cheap ``list_all()`` enumeration.
+    #
+    # Schema version is bumped whenever the entry shape changes; the store
+    # rebuilds the index from the encrypted files on load when the on-disk
+    # version is older.
+    # ------------------------------------------------------------------

-        if index_path.exists():
-            with open(index_path, encoding="utf-8-sig") as f:
-                index = json.load(f)
-        else:
-            index = {"credentials": {}, "version": "1.0"}
+    INDEX_VERSION = "2.0"
+    INDEX_INTERNAL_KEY_NAMES = ("_alias", "_integration_type")

-        if operation == "save":
-            index["credentials"][credential_id] = {
-                "updated_at": datetime.now(UTC).isoformat(),
-                "type": credential_type,
-            }
-        elif operation == "delete":
-            index["credentials"].pop(credential_id, None)
+    def _index_path(self) -> Path:
+        return self.base_path / "metadata" / "index.json"

-        index["last_modified"] = datetime.now(UTC).isoformat()
+    def _read_index(self) -> dict[str, Any]:
+        """Read the index from disk; return an empty skeleton if missing."""
+        path = self._index_path()
+        if not path.exists():
+            return {"version": self.INDEX_VERSION, "credentials": {}}
+        try:
+            with open(path, encoding="utf-8-sig") as f:
+                return json.load(f)
+        except Exception:
+            logger.debug("Failed to read credential index, starting fresh", exc_info=True)
+            return {"version": self.INDEX_VERSION, "credentials": {}}

-        with open(index_path, "w", encoding="utf-8") as f:
-            json.dump(index, f, indent=2)
+    def _write_index(self, index: dict[str, Any]) -> None:
+        """Write the index to disk with consistent envelope fields."""
+        index["version"] = self.INDEX_VERSION
+        index["store_path"] = str(self.base_path)
+        index["generated_at"] = datetime.now(UTC).isoformat()
+        path = self._index_path()
+        path.parent.mkdir(parents=True, exist_ok=True)
+        with open(path, "w", encoding="utf-8") as f:
+            json.dump(index, f, indent=2, sort_keys=False, default=str)
+
+    def _index_entry_for(self, credential: CredentialObject) -> dict[str, Any]:
+        """Build a single index entry from a CredentialObject (no secrets)."""
+        # Visible key names: drop internal markers like _alias / _integration_type
+        # / _identity_* so the entry shows what's actually a credential key.
+        visible_keys = [
+            name
+            for name in credential.keys.keys()
+            if name not in self.INDEX_INTERNAL_KEY_NAMES and not name.startswith("_identity_")
+        ]
+
+        # Earliest expiry across all keys (most likely the access_token).
+        earliest_expiry: datetime | None = None
+        for key in credential.keys.values():
+            if key.expires_at is None:
+                continue
+            if earliest_expiry is None or key.expires_at < earliest_expiry:
+                earliest_expiry = key.expires_at
+
+        return {
+            "credential_type": credential.credential_type.value,
+            "provider": credential.provider_type,
+            "alias": credential.alias,
+            "identity": credential.identity.to_dict(),
+            "key_names": sorted(visible_keys),
+            "created_at": credential.created_at.isoformat() if credential.created_at else None,
+            "updated_at": credential.updated_at.isoformat() if credential.updated_at else None,
+            "last_refreshed": (credential.last_refreshed.isoformat() if credential.last_refreshed else None),
+            "expires_at": earliest_expiry.isoformat() if earliest_expiry else None,
+            "auto_refresh": credential.auto_refresh,
+            "tags": list(credential.tags),
+        }
+
+    def _index_upsert(self, credential: CredentialObject) -> None:
+        """Insert or update one credential entry in the index."""
+        try:
+            index = self._read_index()
+            if index.get("version") != self.INDEX_VERSION:
+                # Old schema — rebuild from disk so we don't blend formats.
+                self._rebuild_index()
+                return
+            credentials = index.setdefault("credentials", {})
+            credentials[credential.id] = self._index_entry_for(credential)
+            self._write_index(index)
+        except Exception:
+            logger.debug("Index upsert failed (non-fatal)", exc_info=True)
+
+    def _index_remove(self, credential_id: str) -> None:
+        """Remove one credential entry from the index."""
+        try:
+            index = self._read_index()
+            if index.get("version") != self.INDEX_VERSION:
+                self._rebuild_index()
+                return
+            credentials = index.setdefault("credentials", {})
+            credentials.pop(credential_id, None)
+            self._write_index(index)
+        except Exception:
+            logger.debug("Index remove failed (non-fatal)", exc_info=True)
+
+    def _maybe_rebuild_index(self) -> None:
+        """Rebuild the index if it's missing, malformed, or on an old schema.
+
+        Called once at startup. The check is cheap — read the version field
+        and bail out if it matches. Encrypted files remain authoritative; this
+        only refreshes the developer-facing snapshot.
+        """
+        path = self._index_path()
+        if path.exists():
+            try:
+                with open(path, encoding="utf-8-sig") as f:
+                    index = json.load(f)
+                if index.get("version") == self.INDEX_VERSION:
+                    return
+            except Exception:
+                pass  # fall through to rebuild
+        self._rebuild_index()
+
+    def _rebuild_index(self) -> None:
+        """Walk the encrypted credentials directory and rewrite a fresh index."""
+        cred_dir = self.base_path / "credentials"
+        if not cred_dir.is_dir():
+            return
+
+        entries: dict[str, Any] = {}
+        for cred_file in sorted(cred_dir.glob("*.enc")):
+            credential_id = cred_file.stem
+            try:
+                cred = self.load(credential_id)
+            except Exception:
+                logger.debug(
+                    "Failed to load %s during index rebuild — skipping",
+                    credential_id,
+                    exc_info=True,
+                )
+                continue
+            if cred is None:
+                continue
+            entries[cred.id] = self._index_entry_for(cred)
+
+        index = {"credentials": entries}
+        self._write_index(index)
+        logger.info("Rebuilt credential index with %d entries", len(entries))


 class EnvVarStorage(CredentialStorage):
@@ -351,8 +474,7 @@ class EnvVarStorage(CredentialStorage):
    def save(self, credential: CredentialObject) -> None:
        """Cannot save to environment variables at runtime."""
        raise NotImplementedError(
-            "EnvVarStorage is read-only. Set environment variables "
-            "externally or use EncryptedFileStorage."
+            "EnvVarStorage is read-only. Set environment variables externally or use EncryptedFileStorage."
        )

    def load(self, credential_id: str) -> CredentialObject | None:
@@ -372,9 +494,7 @@ class EnvVarStorage(CredentialStorage):

    def delete(self, credential_id: str) -> bool:
        """Cannot delete environment variables at runtime."""
-        raise NotImplementedError(
-            "EnvVarStorage is read-only. Unset environment variables externally."
-        )
+        raise NotImplementedError("EnvVarStorage is read-only. Unset environment variables externally.")

    def list_all(self) -> list[str]:
        """List credentials that are available in environment."""
@@ -19,6 +19,7 @@ from typing import Any
 from pydantic import SecretStr

 from .models import (
+    CredentialExpiredError,
    CredentialKey,
    CredentialObject,
    CredentialRefreshError,
@@ -123,9 +124,7 @@ class CredentialStore:
        """
        return self._providers.get(provider_id)

-    def get_provider_for_credential(
-        self, credential: CredentialObject
-    ) -> CredentialProvider | None:
+    def get_provider_for_credential(self, credential: CredentialObject) -> CredentialProvider | None:
        """
        Get the appropriate provider for a credential.

@@ -177,6 +176,8 @@ class CredentialStore:
        self,
        credential_id: str,
        refresh_if_needed: bool = True,
+        *,
+        raise_on_refresh_failure: bool = False,
    ) -> CredentialObject | None:
        """
        Get a credential by ID.
@@ -184,6 +185,11 @@ class CredentialStore:
        Args:
            credential_id: The credential identifier
            refresh_if_needed: If True, refresh expired credentials
+            raise_on_refresh_failure: If True, raise ``CredentialExpiredError``
+                when refresh fails instead of silently returning the stale
+                credential. Tool-execution call sites should pass True so the
+                agent gets a structured "reauth needed" signal rather than a
+                later 401 from the provider.

        Returns:
            CredentialObject or None if not found
@@ -193,7 +199,7 @@ class CredentialStore:
            cached = self._get_from_cache(credential_id)
            if cached is not None:
                if refresh_if_needed and self._should_refresh(cached):
-                    return self._refresh_credential(cached)
+                    return self._refresh_credential(cached, raise_on_failure=raise_on_refresh_failure)
                return cached

            # Load from storage
@@ -203,30 +209,42 @@ class CredentialStore:

            # Refresh if needed
            if refresh_if_needed and self._should_refresh(credential):
-                credential = self._refresh_credential(credential)
+                credential = self._refresh_credential(credential, raise_on_failure=raise_on_refresh_failure)

            # Cache
            self._add_to_cache(credential)

            return credential

-    def get_key(self, credential_id: str, key_name: str) -> str | None:
+    def get_key(
+        self,
+        credential_id: str,
+        key_name: str,
+        *,
+        raise_on_refresh_failure: bool = False,
+    ) -> str | None:
        """
        Convenience method to get a specific key value.

        Args:
            credential_id: The credential identifier
            key_name: The key within the credential
+            raise_on_refresh_failure: See ``get_credential``.

        Returns:
            The key value or None if not found
        """
-        credential = self.get_credential(credential_id)
+        credential = self.get_credential(credential_id, raise_on_refresh_failure=raise_on_refresh_failure)
        if credential is None:
            return None
        return credential.get_key(key_name)

-    def get(self, credential_id: str) -> str | None:
+    def get(
+        self,
+        credential_id: str,
+        *,
+        raise_on_refresh_failure: bool = False,
+    ) -> str | None:
        """
        Legacy compatibility: get the primary key value.

@@ -235,11 +253,12 @@ class CredentialStore:

        Args:
            credential_id: The credential identifier
+            raise_on_refresh_failure: See ``get_credential``.

        Returns:
            The primary key value or None
        """
-        credential = self.get_credential(credential_id)
+        credential = self.get_credential(credential_id, raise_on_refresh_failure=raise_on_refresh_failure)
        if credential is None:
            return None
        return credential.get_default_key()
@@ -510,8 +529,20 @@ class CredentialStore:

        return provider.should_refresh(credential)

-    def _refresh_credential(self, credential: CredentialObject) -> CredentialObject:
-        """Refresh a credential using its provider."""
+    def _refresh_credential(
+        self,
+        credential: CredentialObject,
+        *,
+        raise_on_failure: bool = False,
+    ) -> CredentialObject:
+        """Refresh a credential using its provider.
+
+        When ``raise_on_failure`` is True, a refresh failure raises
+        ``CredentialExpiredError`` carrying provider/alias/help_url metadata
+        for the caller (typically the tool runner) to surface a reauth
+        request. Otherwise, the stale credential is returned to preserve
+        legacy best-effort behavior.
+        """
        provider = self.get_provider_for_credential(credential)
        if provider is None:
            logger.warning(f"No provider found for credential '{credential.id}'")
@@ -530,6 +561,16 @@ class CredentialStore:

        except CredentialRefreshError as e:
            logger.error(f"Failed to refresh credential '{credential.id}': {e}")
+            if raise_on_failure:
+                raise CredentialExpiredError(
+                    credential_id=credential.id,
+                    message=(
+                        f"OAuth token for '{credential.id}' is expired and "
+                        f"refresh failed: {e}. Reauthorization required."
+                    ),
+                    provider=credential.provider_type,
+                    alias=credential.alias,
+                ) from e
            return credential

    def refresh_credential(self, credential_id: str) -> CredentialObject | None:
@@ -88,9 +88,7 @@ class TemplateResolver:
            if key_name:
                value = credential.get_key(key_name)
                if value is None:
-                    raise CredentialKeyNotFoundError(
-                        f"Key '{key_name}' not found in credential '{cred_id}'"
-                    )
+                    raise CredentialKeyNotFoundError(f"Key '{key_name}' not found in credential '{cred_id}'")
            else:
                # Use default key
                value = credential.get_default_key()
@@ -126,9 +124,7 @@ class TemplateResolver:
            ... })
            {"Authorization": "Bearer ghp_xxx", "X-API-Key": "BSAKxxx"}
        """
-        return {
-            key: self.resolve(value, fail_on_missing) for key, value in header_templates.items()
-        }
+        return {key: self.resolve(value, fail_on_missing) for key, value in header_templates.items()}

    def resolve_params(
        self,
@@ -130,9 +130,7 @@ class TestCredentialObject:
        # With access_token
        cred2 = CredentialObject(
            id="test",
-            keys={
-                "access_token": CredentialKey(name="access_token", value=SecretStr("token-value"))
-            },
+            keys={"access_token": CredentialKey(name="access_token", value=SecretStr("token-value"))},
        )
        assert cred2.get_default_key() == "token-value"

@@ -297,9 +295,7 @@ class TestEncryptedFileStorage:
        key = Fernet.generate_key().decode()
        with patch.dict(os.environ, {"HIVE_CREDENTIAL_KEY": key}):
            storage = EncryptedFileStorage(temp_dir)
-            cred = CredentialObject(
-                id="test", keys={"k": CredentialKey(name="k", value=SecretStr("v"))}
-            )
+            cred = CredentialObject(id="test", keys={"k": CredentialKey(name="k", value=SecretStr("v"))})
            storage.save(cred)

            # Create new storage instance with same key
@@ -330,18 +326,10 @@ class TestCompositeStorage:
    def test_read_from_primary(self):
        """Test reading from primary storage."""
        primary = InMemoryStorage()
-        primary.save(
-            CredentialObject(
-                id="test", keys={"k": CredentialKey(name="k", value=SecretStr("primary"))}
-            )
-        )
+        primary.save(CredentialObject(id="test", keys={"k": CredentialKey(name="k", value=SecretStr("primary"))}))

        fallback = InMemoryStorage()
-        fallback.save(
-            CredentialObject(
-                id="test", keys={"k": CredentialKey(name="k", value=SecretStr("fallback"))}
-            )
-        )
+        fallback.save(CredentialObject(id="test", keys={"k": CredentialKey(name="k", value=SecretStr("fallback"))}))

        storage = CompositeStorage(primary, [fallback])
        cred = storage.load("test")
@@ -353,11 +341,7 @@ class TestCompositeStorage:
        """Test fallback when credential not in primary."""
        primary = InMemoryStorage()
        fallback = InMemoryStorage()
-        fallback.save(
-            CredentialObject(
-                id="test", keys={"k": CredentialKey(name="k", value=SecretStr("fallback"))}
-            )
-        )
+        fallback.save(CredentialObject(id="test", keys={"k": CredentialKey(name="k", value=SecretStr("fallback"))}))

        storage = CompositeStorage(primary, [fallback])
        cred = storage.load("test")
@@ -393,9 +377,7 @@ class TestStaticProvider:
    def test_refresh_returns_unchanged(self):
        """Test that refresh returns credential unchanged."""
        provider = StaticProvider()
-        cred = CredentialObject(
-            id="test", keys={"k": CredentialKey(name="k", value=SecretStr("v"))}
-        )
+        cred = CredentialObject(id="test", keys={"k": CredentialKey(name="k", value=SecretStr("v"))})

        refreshed = provider.refresh(cred)
        assert refreshed.get_key("k") == "v"
@@ -403,9 +385,7 @@ class TestStaticProvider:
    def test_validate_with_keys(self):
        """Test validation with keys present."""
        provider = StaticProvider()
-        cred = CredentialObject(
-            id="test", keys={"k": CredentialKey(name="k", value=SecretStr("v"))}
-        )
+        cred = CredentialObject(id="test", keys={"k": CredentialKey(name="k", value=SecretStr("v"))})

        assert provider.validate(cred)

@@ -606,9 +586,7 @@ class TestCredentialStore:
        storage = InMemoryStorage()
        store = CredentialStore(storage=storage, cache_ttl_seconds=60)

-        storage.save(
-            CredentialObject(id="test", keys={"k": CredentialKey(name="k", value=SecretStr("v"))})
-        )
+        storage.save(CredentialObject(id="test", keys={"k": CredentialKey(name="k", value=SecretStr("v"))}))

        # First load
        store.get_credential("test")
@@ -686,9 +664,7 @@ class TestOAuth2Module:
        from core.framework.credentials.oauth2 import OAuth2Config, TokenPlacement

        # Valid config
-        config = OAuth2Config(
-            token_url="https://example.com/token", client_id="id", client_secret="secret"
-        )
+        config = OAuth2Config(token_url="https://example.com/token", client_id="id", client_secret="secret")
        assert config.token_url == "https://example.com/token"

        # Missing token_url
@@ -160,15 +160,9 @@ class CredentialValidationResult:
        if aden_nc:
            if missing or invalid:
                lines.append("")
-            lines.append(
-                "Aden integrations not connected "
-                "(ADEN_API_KEY is set but OAuth tokens unavailable):\n"
-            )
+            lines.append("Aden integrations not connected (ADEN_API_KEY is set but OAuth tokens unavailable):\n")
            for c in aden_nc:
-                lines.append(
-                    f"  {c.env_var} for {_label(c)}"
-                    f"\n    Connect this integration at hive.adenhq.com first."
-                )
+                lines.append(f"  {c.env_var} for {_label(c)}\n    Connect this integration at hive.adenhq.com first.")
        lines.append("\nIf you've already set up credentials, restart your terminal to load them.")
        return "\n".join(lines)

@@ -236,6 +230,45 @@ def _presync_aden_tokens(credential_specs: dict, *, force: bool = False) -> None
            )


+def compute_unavailable_tools(nodes: list) -> tuple[set[str], list[str]]:
+    """Return (tool_names_to_drop, human_messages).
+
+    Runs credential validation *without* raising, collects every tool
+    bound to a failed credential (missing / invalid / Aden-not-connected
+    and no alternative provider available), and returns the set of tool
+    names that should be silently dropped from the worker's effective
+    tool list.
+
+    Use this at every worker-spawn preflight so missing credentials
+    filter tools out of the graph instead of hard-failing the whole
+    spawn. Only affects non-MCP tools — the MCP admission gate
+    (``_build_mcp_admission_gate``) already handles MCP tools at
+    registration time.
+    """
+    try:
+        result = validate_agent_credentials(nodes, verify=False, raise_on_error=False)
+    except Exception as exc:
+        logger.debug("compute_unavailable_tools: validation raised: %s", exc)
+        return set(), []
+
+    drop: set[str] = set()
+    messages: list[str] = []
+    for status in result.failed:
+        if not status.tools:
+            continue
+        drop.update(status.tools)
+        reason = "missing"
+        if status.aden_not_connected:
+            reason = "aden_not_connected"
+        elif status.available and status.valid is False:
+            reason = "invalid"
+        messages.append(
+            f"{status.env_var} ({reason}) → drops {len(status.tools)} tool(s): "
+            f"{', '.join(status.tools[:6])}" + (f" +{len(status.tools) - 6} more" if len(status.tools) > 6 else "")
+        )
+    return drop, messages
+
+
 def validate_agent_credentials(
    nodes: list,
    quiet: bool = False,
@@ -292,9 +325,7 @@ def validate_agent_credentials(
    if os.environ.get("ADEN_API_KEY"):
        _presync_aden_tokens(CREDENTIAL_SPECS, force=force_refresh)

-    env_mapping = {
-        (spec.credential_id or name): spec.env_var for name, spec in CREDENTIAL_SPECS.items()
-    }
+    env_mapping = {(spec.credential_id or name): spec.env_var for name, spec in CREDENTIAL_SPECS.items()}
    env_storage = EnvVarStorage(env_mapping=env_mapping)
    if os.environ.get("HIVE_CREDENTIAL_KEY"):
        storage = CompositeStorage(primary=env_storage, fallbacks=[EncryptedFileStorage()])
@@ -328,12 +359,7 @@ def validate_agent_credentials(
        available = store.is_available(cred_id)

        # Aden-not-connected: ADEN_API_KEY set, Aden-only cred, but integration missing
-        is_aden_nc = (
-            not available
-            and has_aden_key
-            and spec.aden_supported
-            and not spec.direct_api_key_supported
-        )
+        is_aden_nc = not available and has_aden_key and spec.aden_supported and not spec.direct_api_key_supported

        status = CredentialStatus(
            credential_name=cred_name,
@@ -451,9 +477,7 @@ def validate_agent_credentials(
                        identity_data = result.details.get("identity")
                        if identity_data and isinstance(identity_data, dict):
                            try:
-                                cred_obj = store.get_credential(
-                                    status.credential_id, refresh_if_needed=False
-                                )
+                                cred_obj = store.get_credential(status.credential_id, refresh_if_needed=False)
                                if cred_obj:
                                    cred_obj.set_identity(**identity_data)
                                    store.save_credential(cred_obj)
@@ -1,65 +0,0 @@
-"""Graph structures: Goals, Nodes, Edges, and Execution."""
-
-from framework.graph.context import GraphContext
-from framework.graph.context_handoff import ContextHandoff, HandoffContext
-from framework.graph.conversation import ConversationStore, Message, NodeConversation
-from framework.graph.edge import DEFAULT_MAX_TOKENS, EdgeCondition, EdgeSpec, GraphSpec
-from framework.graph.event_loop_node import (
-    EventLoopNode,
-    JudgeProtocol,
-    JudgeVerdict,
-    LoopConfig,
-    OutputAccumulator,
-)
-from framework.graph.executor import GraphExecutor
-from framework.graph.goal import Constraint, Goal, GoalStatus, SuccessCriterion
-from framework.graph.node import NodeContext, NodeProtocol, NodeResult, NodeSpec
-from framework.graph.worker_agent import (
-    Activation,
-    FanOutTag,
-    FanOutTracker,
-    WorkerAgent,
-    WorkerCompletion,
-    WorkerLifecycle,
-)
-
-__all__ = [
-    # Goal
-    "Goal",
-    "SuccessCriterion",
-    "Constraint",
-    "GoalStatus",
-    # Node
-    "NodeSpec",
-    "NodeContext",
-    "NodeResult",
-    "NodeProtocol",
-    # Edge
-    "EdgeSpec",
-    "EdgeCondition",
-    "GraphSpec",
-    "DEFAULT_MAX_TOKENS",
-    # Executor
-    "GraphExecutor",
-    # Conversation
-    "NodeConversation",
-    "ConversationStore",
-    "Message",
-    # Event Loop
-    "EventLoopNode",
-    "LoopConfig",
-    "OutputAccumulator",
-    "JudgeProtocol",
-    "JudgeVerdict",
-    # Context Handoff
-    "ContextHandoff",
-    "HandoffContext",
-    # Worker Agent
-    "WorkerAgent",
-    "WorkerLifecycle",
-    "WorkerCompletion",
-    "Activation",
-    "FanOutTag",
-    "FanOutTracker",
-    "GraphContext",
-]
@@ -1,6 +0,0 @@
-"""EventLoopNode subpackage — modular components of the event loop orchestrator.
-
-All public symbols are re-exported by the parent ``event_loop_node.py`` for
-backward compatibility.  Internal consumers may import directly from these
-submodules for clarity.
-"""
@@ -1,381 +0,0 @@
-"""Subagent execution for the event loop.
-
-Handles the full subagent lifecycle: validation, context setup, tool filtering,
-conversation store derivation, execution, and cleanup.
-"""
-
-from __future__ import annotations
-
-import asyncio
-import json
-import logging
-import time
-from collections.abc import Awaitable, Callable
-from pathlib import Path
-from typing import TYPE_CHECKING, Any
-
-from framework.graph.conversation import ConversationStore
-from framework.graph.event_loop.judge_pipeline import SubagentJudge
-from framework.graph.event_loop.types import LoopConfig, OutputAccumulator
-from framework.graph.node import DataBuffer, NodeContext
-from framework.llm.provider import ToolResult, ToolUse
-from framework.runtime.event_bus import EventBus
-
-if TYPE_CHECKING:
-    from framework.graph.event_loop_node import EventLoopNode
-
-logger = logging.getLogger(__name__)
-
-
-async def execute_subagent(
-    ctx: NodeContext,
-    agent_id: str,
-    task: str,
-    *,
-    config: LoopConfig,
-    event_loop_node_cls: type[EventLoopNode],
-    escalation_receiver_cls: Callable[[], Any],
-    accumulator: OutputAccumulator | None = None,
-    event_bus: EventBus | None = None,
-    tool_executor: Callable[[ToolUse], ToolResult | Awaitable[ToolResult]] | None = None,
-    conversation_store: ConversationStore | None = None,
-    subagent_instance_counter: dict[str, int] | None = None,
-) -> ToolResult:
-    """Execute a subagent and return the result as a ToolResult.
-
-    The subagent:
-    - Gets a fresh conversation with just the task
-    - Has read-only access to the parent's readable memory
-    - Cannot delegate to its own subagents (prevents recursion)
-    - Returns its output in structured JSON format
-
-    Args:
-        ctx: Parent node's context (for memory, tools, LLM access).
-        agent_id: The node ID of the subagent to invoke.
-        task: The task description to give the subagent.
-        accumulator: Parent's OutputAccumulator.
-        event_bus: EventBus for lifecycle events.
-        config: LoopConfig for iteration/tool limits.
-        tool_executor: Tool executor callable.
-        conversation_store: Parent conversation store (for deriving subagent store).
-        subagent_instance_counter: Mutable counter dict for unique subagent paths.
-
-    Returns:
-        ToolResult with structured JSON output.
-    """
-    # Log subagent invocation start
-    logger.info(
-        "\n" + "=" * 60 + "\n"
-        "🤖 SUBAGENT INVOCATION\n"
-        "=" * 60 + "\n"
-        "Parent Node: %s\n"
-        "Subagent ID: %s\n"
-        "Task: %s\n" + "=" * 60,
-        ctx.node_id,
-        agent_id,
-        task[:500] + "..." if len(task) > 500 else task,
-    )
-
-    # 1. Validate agent exists in registry
-    if agent_id not in ctx.node_registry:
-        return ToolResult(
-            tool_use_id="",
-            content=json.dumps(
-                {
-                    "message": f"Sub-agent '{agent_id}' not found in registry",
-                    "data": None,
-                    "metadata": {"agent_id": agent_id, "success": False, "error": "not_found"},
-                }
-            ),
-            is_error=True,
-        )
-
-    subagent_spec = ctx.node_registry[agent_id]
-
-    # 2. Create read-only memory snapshot
-    parent_data = ctx.buffer.read_all()
-
-    # Merge in-flight outputs from the parent's accumulator.
-    if accumulator:
-        for key, value in accumulator.to_dict().items():
-            if key not in parent_data:
-                parent_data[key] = value
-
-    subagent_buffer = DataBuffer()
-    for key, value in parent_data.items():
-        subagent_buffer.write(key, value, validate=False)
-
-    read_keys = set(parent_data.keys()) | set(subagent_spec.input_keys or [])
-    scoped_buffer = subagent_buffer.with_permissions(
-        read_keys=list(read_keys),
-        write_keys=[],  # Read-only!
-    )
-
-    # 2b. Compute instance counter early so the callback and child context
-    # share the same stable node_id for this subagent invocation.
-    if subagent_instance_counter is not None:
-        subagent_instance_counter.setdefault(agent_id, 0)
-        subagent_instance_counter[agent_id] += 1
-        subagent_instance = str(subagent_instance_counter[agent_id])
-    else:
-        subagent_instance = "1"
-
-    if subagent_instance == "1":
-        sa_node_id = f"{ctx.node_id}:subagent:{agent_id}"
-    else:
-        sa_node_id = f"{ctx.node_id}:subagent:{agent_id}:{subagent_instance}"
-
-    # 2c. Set up report callback (one-way channel to parent / event bus)
-    subagent_reports: list[dict] = []
-
-    async def _report_callback(
-        message: str,
-        data: dict | None = None,
-        *,
-        wait_for_response: bool = False,
-    ) -> str | None:
-        subagent_reports.append({"message": message, "data": data, "timestamp": time.time()})
-        if event_bus:
-            await event_bus.emit_subagent_report(
-                stream_id=ctx.node_id,
-                node_id=sa_node_id,
-                subagent_id=agent_id,
-                message=message,
-                data=data,
-                execution_id=ctx.execution_id,
-            )
-
-        if not wait_for_response:
-            return None
-
-        if not event_bus:
-            logger.warning(
-                "Subagent '%s' requested user response but no event_bus available",
-                agent_id,
-            )
-            return None
-
-        # Create isolated receiver and register for input routing
-        import uuid
-
-        escalation_id = f"{ctx.node_id}:escalation:{uuid.uuid4().hex[:8]}"
-        receiver = escalation_receiver_cls()
-        registry = ctx.shared_node_registry
-
-        registry[escalation_id] = receiver
-        try:
-            await event_bus.emit_escalation_requested(
-                stream_id=ctx.stream_id or ctx.node_id,
-                node_id=escalation_id,
-                reason=f"Subagent report (wait_for_response) from {agent_id}",
-                context=message,
-                execution_id=ctx.execution_id,
-            )
-            # Block until queen responds
-            return await receiver.wait()
-        finally:
-            registry.pop(escalation_id, None)
-
-    # 3. Filter tools for subagent
-    subagent_tool_names = set(subagent_spec.tools or [])
-    tool_source = ctx.all_tools if ctx.all_tools else ctx.available_tools
-
-    # GCU auto-population
-    if subagent_spec.node_type == "gcu" and not subagent_tool_names:
-        subagent_tools = [t for t in tool_source if t.name != "delegate_to_sub_agent"]
-    else:
-        subagent_tools = [
-            t
-            for t in tool_source
-            if t.name in subagent_tool_names and t.name != "delegate_to_sub_agent"
-        ]
-
-    missing = subagent_tool_names - {t.name for t in subagent_tools}
-    if missing:
-        logger.warning(
-            "Subagent '%s' requested tools not found in catalog: %s",
-            agent_id,
-            sorted(missing),
-        )
-
-    logger.info(
-        "📦 Subagent '%s' configuration:\n"
-        "   - System prompt: %s\n"
-        "   - Tools available (%d): %s\n"
-        "   - Memory keys inherited: %s",
-        agent_id,
-        (subagent_spec.system_prompt[:200] + "...")
-        if subagent_spec.system_prompt and len(subagent_spec.system_prompt) > 200
-        else subagent_spec.system_prompt,
-        len(subagent_tools),
-        [t.name for t in subagent_tools],
-        list(parent_data.keys()),
-    )
-
-    # 4. Build subagent context
-    max_iter = min(config.max_iterations, 10)
-    subagent_ctx = NodeContext(
-        runtime=ctx.runtime,
-        node_id=sa_node_id,
-        node_spec=subagent_spec,
-        buffer=scoped_buffer,
-        input_data={"task": task, **parent_data},
-        llm=ctx.llm,
-        available_tools=subagent_tools,
-        goal_context=(
-            f"Your specific task: {task}\n\n"
-            f"COMPLETION REQUIREMENTS:\n"
-            f"When your task is done, you MUST call set_output() "
-            f"for each required key: {subagent_spec.output_keys}\n"
-            f"Alternatively, call report_to_parent(mark_complete=true) "
-            f"with your findings in message/data.\n"
-            f"You have a maximum of {max_iter} turns to complete this task."
-        ),
-        goal=ctx.goal,
-        max_tokens=ctx.max_tokens,
-        runtime_logger=ctx.runtime_logger,
-        is_subagent_mode=True,  # Prevents nested delegation
-        report_callback=_report_callback,
-        node_registry={},  # Empty - no nested subagents
-        shared_node_registry=ctx.shared_node_registry,  # For escalation routing
-    )
-
-    # 5. Create and execute subagent EventLoopNode
-    subagent_conv_store = None
-    if conversation_store is not None:
-        from framework.storage.conversation_store import FileConversationStore
-
-        parent_base = getattr(conversation_store, "_base", None)
-        if parent_base is not None:
-            conversations_dir = parent_base.parent
-            subagent_dir_name = f"{agent_id}-{subagent_instance}"
-            subagent_store_path = conversations_dir / subagent_dir_name
-            subagent_conv_store = FileConversationStore(base_path=subagent_store_path)
-
-    # Derive a subagent-scoped spillover dir
-    subagent_spillover = None
-    if config.spillover_dir:
-        subagent_spillover = str(Path(config.spillover_dir) / agent_id / subagent_instance)
-
-    subagent_node = event_loop_node_cls(
-        event_bus=event_bus,
-        judge=SubagentJudge(task=task, max_iterations=max_iter),
-        config=LoopConfig(
-            max_iterations=max_iter,
-            max_tool_calls_per_turn=config.max_tool_calls_per_turn,
-            tool_call_overflow_margin=config.tool_call_overflow_margin,
-            max_context_tokens=config.max_context_tokens,
-            stall_detection_threshold=config.stall_detection_threshold,
-            max_tool_result_chars=config.max_tool_result_chars,
-            spillover_dir=subagent_spillover,
-        ),
-        tool_executor=tool_executor,
-        conversation_store=subagent_conv_store,
-    )
-
-    # Each subagent instance gets its own unique browser profile so concurrent
-    # subagents don't share tab groups. The profile is injected into every
-    # browser_* tool call by wrapping the tool executor.
-    _gcu_profile = f"{agent_id}:{subagent_instance}"
-    _original_tool_executor = None
-
-    if tool_executor is not None:
-        _original_tool_executor = tool_executor
-
-        async def _gcu_profile_injecting_executor(
-            tool_use: ToolUse,
-        ) -> ToolResult | Awaitable[ToolResult]:
-            if tool_use.name.startswith("browser_") and "profile" not in (tool_use.input or {}):
-                from dataclasses import replace
-
-                tool_use = replace(tool_use, input={**(tool_use.input or {}), "profile": _gcu_profile})
-            result = _original_tool_executor(tool_use)
-            if asyncio.isfuture(result) or asyncio.iscoroutine(result):
-                return await result
-            return result
-
-        tool_executor = _gcu_profile_injecting_executor
-
-    try:
-        logger.info("🚀 Starting subagent '%s' execution...", agent_id)
-        start_time = time.time()
-        result = await subagent_node.execute(subagent_ctx)
-        latency_ms = int((time.time() - start_time) * 1000)
-
-        separator = "-" * 60
-        logger.info(
-            "\n%s\n"
-            "✅ SUBAGENT '%s' COMPLETED\n"
-            "%s\n"
-            "Success: %s\n"
-            "Latency: %dms\n"
-            "Tokens used: %s\n"
-            "Output keys: %s\n"
-            "%s",
-            separator,
-            agent_id,
-            separator,
-            result.success,
-            latency_ms,
-            result.tokens_used,
-            list(result.output.keys()) if result.output else [],
-            separator,
-        )
-
-        result_json = {
-            "message": (
-                f"Sub-agent '{agent_id}' completed successfully"
-                if result.success
-                else f"Sub-agent '{agent_id}' failed: {result.error}"
-            ),
-            "data": result.output,
-            "reports": subagent_reports if subagent_reports else None,
-            "metadata": {
-                "agent_id": agent_id,
-                "success": result.success,
-                "tokens_used": result.tokens_used,
-                "latency_ms": latency_ms,
-                "report_count": len(subagent_reports),
-            },
-        }
-
-        return ToolResult(
-            tool_use_id="",
-            content=json.dumps(result_json, indent=2, default=str),
-            is_error=not result.success,
-        )
-
-    except Exception as e:
-        logger.exception(
-            "\n" + "!" * 60 + "\n❌ SUBAGENT '%s' FAILED\nError: %s\n" + "!" * 60,
-            agent_id,
-            str(e),
-        )
-        result_json = {
-            "message": f"Sub-agent '{agent_id}' raised exception: {e}",
-            "data": None,
-            "metadata": {
-                "agent_id": agent_id,
-                "success": False,
-                "error": str(e),
-            },
-        }
-        return ToolResult(
-            tool_use_id="",
-            content=json.dumps(result_json, indent=2),
-            is_error=True,
-        )
-    finally:
-        # Close the tab group this subagent created, if any.
-        if _original_tool_executor is not None:
-            try:
-                stop_call = ToolUse(
-                    id="__subagent_cleanup__",
-                    name="browser_stop",
-                    input={"profile": _gcu_profile},
-                )
-                result = _original_tool_executor(stop_call)
-                if asyncio.isfuture(result) or asyncio.iscoroutine(result):
-                    await result
-            except Exception:
-                pass
@@ -1,369 +0,0 @@
-"""Synthetic tool builders for the event loop.
-
-Factory functions that create ``Tool`` definitions for framework-level
-synthetic tools (set_output, ask_user, escalate, delegate, report_to_parent).
-Also includes the ``handle_set_output`` validation logic.
-
-All functions are pure — they receive explicit parameters and return
-``Tool`` or ``ToolResult`` objects with no side effects.
-"""
-
-from __future__ import annotations
-
-from typing import Any
-
-from framework.llm.provider import Tool, ToolResult
-
-
-def build_ask_user_tool() -> Tool:
-    """Build the synthetic ask_user tool for explicit user-input requests.
-
-    The queen calls ask_user() when it needs to pause and wait
-    for user input.  Text-only turns WITHOUT ask_user flow through without
-    blocking, allowing progress updates and summaries to stream freely.
-    """
-    return Tool(
-        name="ask_user",
-        description=(
-            "You MUST call this tool whenever you need the user's response. "
-            "Always call it after greeting the user, asking a question, or "
-            "requesting approval. Do NOT call it for status updates or "
-            "summaries that don't require a response. "
-            "Always include 2-3 predefined options. The UI automatically "
-            "appends an 'Other' free-text input after your options, so NEVER "
-            "include catch-all options like 'Custom idea', 'Something else', "
-            "'Other', or 'None of the above' — the UI handles that. "
-            "When the question primarily needs a typed answer but you must "
-            "include options, make one option signal that typing is expected "
-            "(e.g. 'I\\'ll type my response'). This helps users discover the "
-            "free-text input. "
-            "The ONLY exception: omit options when the question demands a "
-            "free-form answer the user must type out (e.g. 'Describe your "
-            "agent idea', 'Paste the error message'). "
-            '{"question": "What would you like to do?", "options": '
-            '["Build a new agent", "Modify existing agent", "Run tests"]} '
-            "Free-form example: "
-            '{"question": "Describe the agent you want to build."}'
-        ),
-        parameters={
-            "type": "object",
-            "properties": {
-                "question": {
-                    "type": "string",
-                    "description": "The question or prompt shown to the user.",
-                },
-                "options": {
-                    "type": "array",
-                    "items": {"type": "string"},
-                    "description": (
-                        "2-3 specific predefined choices. Include in most cases. "
-                        'Example: ["Option A", "Option B", "Option C"]. '
-                        "The UI always appends an 'Other' free-text input, so "
-                        "do NOT include catch-alls like 'Custom idea' or 'Other'. "
-                        "Omit ONLY when the user must type a free-form answer."
-                    ),
-                    "minItems": 2,
-                    "maxItems": 3,
-                },
-            },
-            "required": ["question"],
-        },
-    )
-
-
-def build_ask_user_multiple_tool() -> Tool:
-    """Build the synthetic ask_user_multiple tool for batched questions.
-
-    Queen-only tool that presents multiple questions at once so the user
-    can answer them all in a single interaction rather than one at a time.
-    """
-    return Tool(
-        name="ask_user_multiple",
-        description=(
-            "Ask the user multiple questions at once. Use this instead of "
-            "ask_user when you have 2 or more questions to ask in the same "
-            "turn — it lets the user answer everything in one go rather than "
-            "going back and forth. Each question can have its own predefined "
-            "options (2-3 choices) or be free-form. The UI renders all "
-            "questions together with a single Submit button. "
-            "ALWAYS prefer this over ask_user when you have multiple things "
-            "to clarify. "
-            "IMPORTANT: Do NOT repeat the questions in your text response — "
-            "the widget renders them. Keep your text to a brief intro only. "
-            '{"questions": ['
-            '  {"id": "scope", "prompt": "What scope?", "options": ["Full", "Partial"]},'
-            '  {"id": "format", "prompt": "Output format?", "options": ["PDF", "CSV", "JSON"]},'
-            '  {"id": "details", "prompt": "Any special requirements?"}'
-            "]}"
-        ),
-        parameters={
-            "type": "object",
-            "properties": {
-                "questions": {
-                    "type": "array",
-                    "items": {
-                        "type": "object",
-                        "properties": {
-                            "id": {
-                                "type": "string",
-                                "description": (
-                                    "Short identifier for this question (used in the response)."
-                                ),
-                            },
-                            "prompt": {
-                                "type": "string",
-                                "description": "The question text shown to the user.",
-                            },
-                            "options": {
-                                "type": "array",
-                                "items": {"type": "string"},
-                                "description": (
-                                    "2-3 predefined choices. The UI appends an "
-                                    "'Other' free-text input automatically. "
-                                    "Omit only when the user must type a free-form answer."
-                                ),
-                                "minItems": 2,
-                                "maxItems": 3,
-                            },
-                        },
-                        "required": ["id", "prompt"],
-                    },
-                    "minItems": 2,
-                    "maxItems": 8,
-                    "description": "List of questions to present to the user.",
-                },
-            },
-            "required": ["questions"],
-        },
-    )
-
-
-def build_set_output_tool(output_keys: list[str] | None) -> Tool | None:
-    """Build the synthetic set_output tool for explicit output declaration."""
-    if not output_keys:
-        return None
-    return Tool(
-        name="set_output",
-        description=(
-            "Set an output value for this node. Call once per output key. "
-            "Use this for brief notes, counts, status, and file references — "
-            "NOT for large data payloads. When a tool result was saved to a "
-            "data file, pass the filename as the value "
-            "(e.g. 'google_sheets_get_values_1.txt') so the next phase can "
-            "load the full data. Values exceeding ~2000 characters are "
-            "auto-saved to data files. "
-            f"Valid keys: {output_keys}"
-        ),
-        parameters={
-            "type": "object",
-            "properties": {
-                "key": {
-                    "type": "string",
-                    "description": f"Output key. Must be one of: {output_keys}",
-                    "enum": output_keys,
-                },
-                "value": {
-                    "type": "string",
-                    "description": (
-                        "The output value — a brief note, count, status, "
-                        "or data filename reference."
-                    ),
-                },
-            },
-            "required": ["key", "value"],
-        },
-    )
-
-
-def build_escalate_tool() -> Tool:
-    """Build the synthetic escalate tool for worker -> queen handoff."""
-    return Tool(
-        name="escalate",
-        description=(
-            "Escalate to the queen when requesting user input, "
-            "blocked by errors, missing "
-            "credentials, or ambiguous constraints that require supervisor "
-            "guidance. Include a concise reason and optional context. "
-            "The node will pause until the queen injects guidance."
-        ),
-        parameters={
-            "type": "object",
-            "properties": {
-                "reason": {
-                    "type": "string",
-                    "description": (
-                        "Short reason for escalation (e.g. 'Tool repeatedly failing')."
-                    ),
-                },
-                "context": {
-                    "type": "string",
-                    "description": "Optional diagnostic details for the queen.",
-                },
-            },
-            "required": ["reason"],
-        },
-    )
-
-
-def build_delegate_tool(sub_agents: list[str], node_registry: dict[str, Any]) -> Tool | None:
-    """Build the synthetic delegate_to_sub_agent tool for subagent invocation.
-
-    Args:
-        sub_agents: List of node IDs that can be invoked as subagents.
-        node_registry: Map of node_id -> NodeSpec for looking up subagent descriptions.
-
-    Returns:
-        Tool definition if sub_agents is non-empty, None otherwise.
-    """
-    if not sub_agents:
-        return None
-
-    agent_descriptions = []
-    for agent_id in sub_agents:
-        spec = node_registry.get(agent_id)
-        if spec:
-            desc = getattr(spec, "description", "(no description)")
-            agent_descriptions.append(f"- {agent_id}: {desc}")
-        else:
-            agent_descriptions.append(f"- {agent_id}: (not found in registry)")
-
-    return Tool(
-        name="delegate_to_sub_agent",
-        description=(
-            "Delegate a task to a specialized sub-agent. The sub-agent runs "
-            "autonomously with read-only access to current memory and returns "
-            "its result. Use this to parallelize work or leverage specialized capabilities.\n\n"
-            "Available sub-agents:\n" + "\n".join(agent_descriptions)
-        ),
-        parameters={
-            "type": "object",
-            "properties": {
-                "agent_id": {
-                    "type": "string",
-                    "description": f"The sub-agent to invoke. Must be one of: {sub_agents}",
-                    "enum": sub_agents,
-                },
-                "task": {
-                    "type": "string",
-                    "description": (
-                        "The task description for the sub-agent to execute. "
-                        "Be specific about what you want the sub-agent to do and "
-                        "what information to return."
-                    ),
-                },
-            },
-            "required": ["agent_id", "task"],
-        },
-    )
-
-
-def build_report_to_parent_tool() -> Tool:
-    """Build the synthetic report_to_parent tool for sub-agent progress reports.
-
-    Sub-agents call this to send one-way progress updates, partial findings,
-    or status reports to the parent node (and external observers via event bus)
-    without blocking execution.
-
-    When ``wait_for_response`` is True, the sub-agent blocks until the parent
-    relays the user's response — used for escalation (e.g. login pages, CAPTCHAs).
-
-    When ``mark_complete`` is True, the sub-agent terminates immediately after
-    sending the report — no need to call set_output for each output key.
-    """
-    return Tool(
-        name="report_to_parent",
-        description=(
-            "Send a report to the parent agent. By default this is fire-and-forget: "
-            "the parent receives the report but does not respond. "
-            "Set wait_for_response=true to BLOCK until the user replies — use this "
-            "when you need human intervention (e.g. login pages, CAPTCHAs, "
-            "authentication walls). The user's response is returned as the tool result. "
-            "Set mark_complete=true to finish your task and terminate immediately "
-            "after sending the report — use this when your findings are in the "
-            "message/data fields and you don't need to call set_output."
-        ),
-        parameters={
-            "type": "object",
-            "properties": {
-                "message": {
-                    "type": "string",
-                    "description": "A human-readable status or progress message.",
-                },
-                "data": {
-                    "type": "object",
-                    "description": "Optional structured data to include with the report.",
-                },
-                "wait_for_response": {
-                    "type": "boolean",
-                    "description": (
-                        "If true, block execution until the user responds. "
-                        "Use for escalation scenarios requiring human intervention."
-                    ),
-                    "default": False,
-                },
-                "mark_complete": {
-                    "type": "boolean",
-                    "description": (
-                        "If true, terminate the sub-agent immediately after sending "
-                        "this report. The report message and data are delivered to the "
-                        "parent as the final result. No set_output calls are needed."
-                    ),
-                    "default": False,
-                },
-            },
-            "required": ["message"],
-        },
-    )
-
-
-def handle_set_output(
-    tool_input: dict[str, Any],
-    output_keys: list[str] | None,
-) -> ToolResult:
-    """Handle set_output tool call. Returns ToolResult (sync)."""
-    import logging
-    import re
-
-    logger = logging.getLogger(__name__)
-
-    key = tool_input.get("key", "")
-    value = tool_input.get("value", "")
-    valid_keys = output_keys or []
-
-    # Recover from truncated JSON (max_tokens hit mid-argument).
-    # The _raw key is set by litellm when json.loads fails.
-    if not key and "_raw" in tool_input:
-        raw = tool_input["_raw"]
-        key_match = re.search(r'"key"\s*:\s*"(\w+)"', raw)
-        if key_match:
-            key = key_match.group(1)
-        val_match = re.search(r'"value"\s*:\s*"', raw)
-        if val_match:
-            start = val_match.end()
-            value = raw[start:].rstrip()
-            for suffix in ('"}\n', '"}', '"'):
-                if value.endswith(suffix):
-                    value = value[: -len(suffix)]
-                    break
-        if key:
-            logger.warning(
-                "Recovered set_output args from truncated JSON: key=%s, value_len=%d",
-                key,
-                len(value),
-            )
-            # Re-inject so the caller sees proper key/value
-            tool_input["key"] = key
-            tool_input["value"] = value
-
-    if key not in valid_keys:
-        return ToolResult(
-            tool_use_id="",
-            content=f"Invalid output key '{key}'. Valid keys: {valid_keys}",
-            is_error=True,
-        )
-
-    return ToolResult(
-        tool_use_id="",
-        content=f"Output '{key}' set successfully.",
-        is_error=False,
-    )
@@ -1,151 +0,0 @@
-"""Streaming XML tag filter for thinking tags.
-
-Strips configured XML tags (e.g. ``<situation>``, ``<monologue>``) from
-a chunked text stream while preserving the full text for conversation
-storage.  The filter is stateful — it handles chunks that split mid-tag.
-
-Only touches text content.  Tool calls flow through a completely separate
-code path and are never affected by this filter.
-"""
-
-from __future__ import annotations
-
-from collections.abc import Sequence
-
-
-class ThinkingTagFilter:
-    """Strips XML thinking tags from a streaming text output.
-
-    Buffers content inside configured tags and yields only the visible
-    content outside those tags.  Handles chunks that split across tag
-    boundaries (e.g. a chunk ending with ``"<mono"``).
-
-    Args:
-        tag_names: Tag names to strip (e.g. ``["situation", "monologue"]``).
-    """
-
-    def __init__(self, tag_names: Sequence[str]) -> None:
-        self._tag_names: set[str] = set(tag_names)
-        # Pre-compute all opening and closing tag strings for matching.
-        self._open_tags: dict[str, str] = {name: f"<{name}>" for name in tag_names}
-        self._close_tags: dict[str, str] = {name: f"</{name}>" for name in tag_names}
-        # All possible tag prefixes for partial-match detection.
-        self._all_tag_strings: list[str] = sorted(
-            list(self._open_tags.values()) + list(self._close_tags.values()),
-            key=len,
-            reverse=True,
-        )
-
-        self._inside_tag: str | None = None  # Which tag we're inside, or None.
-        self._pending: str = ""  # Chars that might be a partial tag.
-        self._visible_text: str = ""  # Accumulated visible snapshot.
-
-    def feed(self, chunk: str) -> str:
-        """Feed a text chunk and return the visible portion.
-
-        Characters inside thinking tags are suppressed.  Characters that
-        *might* be the start of a tag are buffered until the next chunk
-        resolves the ambiguity.
-
-        Returns:
-            The portion of text that should be shown to the user.
-        """
-        buf = self._pending + chunk
-        self._pending = ""
-        visible = self._process(buf)
-        self._visible_text += visible
-        return visible
-
-    @property
-    def visible_snapshot(self) -> str:
-        """Accumulated visible text so far (for the snapshot field)."""
-        return self._visible_text
-
-    def flush(self) -> str:
-        """Flush any pending partial tag as visible text.
-
-        Called at end-of-stream.  If characters were buffered because they
-        looked like the start of a tag but the stream ended before the tag
-        completed, they are emitted as visible text (graceful degradation).
-        """
-        result = ""
-        if self._pending:
-            if self._inside_tag is None:
-                result = self._pending
-            # If inside a tag, discard pending (unclosed tag content).
-            self._pending = ""
-        self._visible_text += result
-        return result
-
-    # ------------------------------------------------------------------
-    # Internal processing
-    # ------------------------------------------------------------------
-
-    def _process(self, buf: str) -> str:
-        """Process a buffer, returning visible text and updating state."""
-        visible_parts: list[str] = []
-        i = 0
-        n = len(buf)
-
-        while i < n:
-            if self._inside_tag is not None:
-                # Inside a tag — look for the closing tag.
-                close = self._close_tags[self._inside_tag]
-                close_pos = buf.find(close, i)
-                if close_pos == -1:
-                    # Closing tag might be split across chunks.
-                    # Check if the tail of buf is a prefix of the close tag.
-                    tail_len = min(len(close) - 1, n - i)
-                    for tl in range(tail_len, 0, -1):
-                        if close.startswith(buf[n - tl :]):
-                            self._pending = buf[n - tl :]
-                            i = n
-                            break
-                    else:
-                        # No partial match — discard everything (inside tag).
-                        i = n
-                    break
-                else:
-                    # Found closing tag — skip past it and exit tag.
-                    i = close_pos + len(close)
-                    self._inside_tag = None
-            else:
-                # Outside any tag — look for '<'.
-                lt_pos = buf.find("<", i)
-                if lt_pos == -1:
-                    # No '<' — everything is visible.
-                    visible_parts.append(buf[i:])
-                    i = n
-                else:
-                    # Emit text before the '<'.
-                    if lt_pos > i:
-                        visible_parts.append(buf[i:lt_pos])
-                    # Try to match an opening tag at this position.
-                    remainder = buf[lt_pos:]
-                    matched = False
-                    for name, open_tag in self._open_tags.items():
-                        if remainder.startswith(open_tag):
-                            # Full opening tag found — enter tag.
-                            self._inside_tag = name
-                            i = lt_pos + len(open_tag)
-                            matched = True
-                            break
-                    if not matched:
-                        # Check if remainder could be a partial tag prefix.
-                        if self._is_partial_tag_prefix(remainder):
-                            # Buffer and wait for next chunk.
-                            self._pending = remainder
-                            i = n
-                        else:
-                            # Not a known tag — '<' is visible text.
-                            visible_parts.append("<")
-                            i = lt_pos + 1
-
-        return "".join(visible_parts)
-
-    def _is_partial_tag_prefix(self, text: str) -> bool:
-        """Check if text could be the start of a known tag string."""
-        for tag_str in self._all_tag_strings:
-            if tag_str.startswith(text) and len(text) < len(tag_str):
-                return True
-        return False
@@ -1,207 +0,0 @@
-"""Shared types and state containers for the event loop package."""
-
-from __future__ import annotations
-
-import json
-import logging
-import time
-from dataclasses import dataclass, field
-from pathlib import Path
-from typing import Any, Literal, Protocol, runtime_checkable
-
-from framework.graph.conversation import (
-    ConversationStore,
-    get_run_cursor,
-    update_run_cursor,
-)
-
-logger = logging.getLogger(__name__)
-
-
-@dataclass
-class TriggerEvent:
-    """A framework-level trigger signal (timer tick or webhook hit)."""
-
-    trigger_type: str
-    source_id: str
-    payload: dict[str, Any] = field(default_factory=dict)
-    timestamp: float = field(default_factory=time.time)
-
-
-@dataclass
-class JudgeVerdict:
-    """Result of judge evaluation for the event loop."""
-
-    action: Literal["ACCEPT", "RETRY", "ESCALATE"]
-    # None  = no evaluation happened (skip_judge, tool-continue); not logged.
-    # ""    = evaluated but no feedback; logged with default text.
-    # "..." = evaluated with feedback; logged as-is.
-    feedback: str | None = None
-
-
-@runtime_checkable
-class JudgeProtocol(Protocol):
-    """Protocol for event-loop judges."""
-
-    async def evaluate(self, context: dict[str, Any]) -> JudgeVerdict: ...
-
-
-@dataclass
-class LoopConfig:
-    """Configuration for the event loop."""
-
-    max_iterations: int = 50
-    max_tool_calls_per_turn: int = 30
-    judge_every_n_turns: int = 1
-    stall_detection_threshold: int = 3
-    stall_similarity_threshold: float = 0.85
-    max_context_tokens: int = 32_000
-    store_prefix: str = ""
-
-    # Overflow margin for max_tool_calls_per_turn. Tool calls are only
-    # discarded when the count exceeds max_tool_calls_per_turn * (1 + margin).
-    tool_call_overflow_margin: float = 0.5
-
-    # Tool result context management.
-    max_tool_result_chars: int = 30_000
-    spillover_dir: str | None = None
-
-    # set_output value spilling.
-    max_output_value_chars: int = 2_000
-
-    # Stream retry.
-    max_stream_retries: int = 3
-    stream_retry_backoff_base: float = 2.0
-    stream_retry_max_delay: float = 60.0
-
-    # Tool doom loop detection.
-    tool_doom_loop_threshold: int = 3
-
-    # Client-facing auto-block grace period.
-    cf_grace_turns: int = 1
-    # Worker auto-escalation: text-only turns before escalating to queen.
-    worker_escalation_grace_turns: int = 1
-    tool_doom_loop_enabled: bool = True
-
-    # Per-tool-call timeout.
-    tool_call_timeout_seconds: float = 60.0
-
-    # Subagent delegation timeout (wall-clock max).
-    subagent_timeout_seconds: float = 3600.0
-
-    # Subagent inactivity timeout - only timeout if no activity for this duration.
-    # This resets whenever the subagent makes progress (tool calls, LLM responses).
-    # Set to 0 to use only the wall-clock timeout.
-    subagent_inactivity_timeout_seconds: float = 300.0
-
-    # Lifecycle hooks.
-    hooks: dict[str, list] | None = None
-
-    def __post_init__(self) -> None:
-        if self.hooks is None:
-            object.__setattr__(self, "hooks", {})
-
-
-@dataclass
-class HookContext:
-    """Context passed to every lifecycle hook."""
-
-    event: str
-    trigger: str | None
-    system_prompt: str
-
-
-@dataclass
-class HookResult:
-    """What a hook may return to modify node state."""
-
-    system_prompt: str | None = None
-    inject: str | None = None
-
-
-@dataclass
-class OutputAccumulator:
-    """Accumulates output key-value pairs with optional write-through persistence."""
-
-    values: dict[str, Any] = field(default_factory=dict)
-    store: ConversationStore | None = None
-    spillover_dir: str | None = None
-    max_value_chars: int = 0
-    run_id: str | None = None
-
-    async def set(self, key: str, value: Any) -> None:
-        """Set a key-value pair, auto-spilling large values to files."""
-        value = self._auto_spill(key, value)
-        self.values[key] = value
-        if self.store:
-            cursor = await self.store.read_cursor() or {}
-            outputs = cursor.get("outputs", {})
-            outputs[key] = value
-            cursor["outputs"] = outputs
-            await self.store.write_cursor(cursor)
-
-    def _auto_spill(self, key: str, value: Any) -> Any:
-        """Save large values to a file and return a reference string."""
-        if self.max_value_chars <= 0 or not self.spillover_dir:
-            return value
-
-        val_str = json.dumps(value, ensure_ascii=False) if not isinstance(value, str) else value
-        if len(val_str) <= self.max_value_chars:
-            return value
-
-        spill_path = Path(self.spillover_dir)
-        spill_path.mkdir(parents=True, exist_ok=True)
-        ext = ".json" if isinstance(value, (dict, list)) else ".txt"
-        filename = f"output_{key}{ext}"
-        write_content = (
-            json.dumps(value, indent=2, ensure_ascii=False)
-            if isinstance(value, (dict, list))
-            else str(value)
-        )
-        file_path = spill_path / filename
-        file_path.write_text(write_content, encoding="utf-8")
-        file_size = file_path.stat().st_size
-        logger.info(
-            "set_output value auto-spilled: key=%s, %d chars -> %s (%d bytes)",
-            key,
-            len(val_str),
-            filename,
-            file_size,
-        )
-        # Use absolute path so parent agents can find files from subagents
-        abs_path = str(file_path.resolve())
-        return (
-            f"[Saved to '{abs_path}' ({file_size:,} bytes). "
-            f"Use read_file(path='{abs_path}') "
-            f"to access full data.]"
-        )
-
-    def get(self, key: str) -> Any | None:
-        return self.values.get(key)
-
-    def to_dict(self) -> dict[str, Any]:
-        return dict(self.values)
-
-    def has_all_keys(self, required: list[str]) -> bool:
-        return all(key in self.values and self.values[key] is not None for key in required)
-
-    @classmethod
-    async def restore(
-        cls,
-        store: ConversationStore,
-        run_id: str | None = None,
-    ) -> OutputAccumulator:
-        cursor = await store.read_cursor()
-        values = cursor.get("outputs", {}) if cursor else {}
-        return cls(values=values, store=store, run_id=run_id)
-
-
-__all__ = [
-    "HookContext",
-    "HookResult",
-    "JudgeProtocol",
-    "JudgeVerdict",
-    "LoopConfig",
-    "OutputAccumulator",
-    "TriggerEvent",
-]
@@ -1,129 +0,0 @@
-"""GCU (browser automation) node type constants.
-
-A ``gcu`` node is an ``event_loop`` node with two automatic enhancements:
-1. A canonical browser best-practices system prompt is prepended.
-2. All tools from the GCU MCP server are auto-included.
-
-No new ``NodeProtocol`` subclass — the ``gcu`` type is purely a declarative
-signal processed by the runner and executor at setup time.
-"""
-
-# ---------------------------------------------------------------------------
-# MCP server identity
-# ---------------------------------------------------------------------------
-
-GCU_SERVER_NAME = "gcu-tools"
-"""Name used to identify the GCU MCP server in ``mcp_servers.json``."""
-
-GCU_MCP_SERVER_CONFIG: dict = {
-    "name": GCU_SERVER_NAME,
-    "transport": "stdio",
-    "command": "uv",
-    "args": ["run", "python", "-m", "gcu.server", "--stdio"],
-    "cwd": "../../tools",
-    "description": "GCU tools for browser automation",
-}
-"""Default stdio config for the GCU MCP server (relative to exports/<agent>/)."""
-
-# ---------------------------------------------------------------------------
-# Browser best-practices system prompt
-# ---------------------------------------------------------------------------
-
-GCU_BROWSER_SYSTEM_PROMPT = """\
-# Browser Automation Best Practices
-
-Follow these rules for reliable, efficient browser interaction.
-
-## Reading Pages
- ALWAYS prefer `browser_snapshot` over `browser_get_text("body")`
-  — it returns a compact ~1-5 KB accessibility tree vs 100+ KB of raw HTML.
- Interaction tools (`browser_click`, `browser_type`, `browser_fill`,
-  `browser_scroll`, etc.) return a page snapshot automatically in their
-  result. Use it to decide your next action — do NOT call
-  `browser_snapshot` separately after every action.
-  Only call `browser_snapshot` when you need a fresh view without
-  performing an action, or after setting `auto_snapshot=false`.
- Do NOT use `browser_screenshot` to read text — use
-  `browser_snapshot` for that (compact, searchable, fast).
- DO use `browser_screenshot` when you need visual context:
-  charts, images, canvas elements, layout verification, or when
-  the snapshot doesn't capture what you need.
- Only fall back to `browser_get_text` for extracting specific
-  small elements by CSS selector.
-
-## Navigation & Waiting
- `browser_navigate` and `browser_open` already wait for the page to
-  load (`domcontentloaded`). Do NOT call `browser_wait` with no
-  arguments after navigation — it wastes time.
-  Only use `browser_wait` when you need a *specific element* or *text*
-  to appear (pass `selector` or `text`).
- NEVER re-navigate to the same URL after scrolling
-  — this resets your scroll position and loses loaded content.
-
-## Scrolling
- Use large scroll amounts ~2000 when loading more content
-  — sites like twitter and linkedin have lazy loading for paging.
- The scroll result includes a snapshot automatically — no need to call
-  `browser_snapshot` separately.
-
-## Batching Actions
- You can call multiple tools in a single turn — they execute in parallel.
-  ALWAYS batch independent actions together. Examples:
-  - Fill multiple form fields in one turn.
-  - Navigate + snapshot in one turn.
-  - Click + scroll if targeting different elements.
- When batching, set `auto_snapshot=false` on all but the last action
-  to avoid redundant snapshots.
- Aim for 3-5 tool calls per turn minimum. One tool call per turn is
-  wasteful.
-
-## Error Recovery
- If a tool fails, retry once with the same approach.
- If it fails a second time, STOP retrying and switch approach.
- If `browser_snapshot` fails → try `browser_get_text` with a
-  specific small selector as fallback.
- If `browser_open` fails or page seems stale → `browser_stop`,
-  then `browser_start`, then retry.
-
-## Tab Management
-
-**Close tabs as soon as you are done with them** — not only at the end of the task.
-After reading or extracting data from a tab, close it immediately.
-
-**Decision rules:**
- Finished reading/extracting from a tab? → `browser_close(target_id=...)`
- Completed a multi-tab workflow? → `browser_close_finished()` to clean up all your tabs
- More than 3 tabs open? → stop and close finished ones before opening more
- Popup appeared that you didn't need? → close it immediately
-
-**Origin awareness:** `browser_tabs` returns an `origin` field for each tab:
- `"agent"` — you opened it; you own it; close it when done
- `"popup"` — opened by a link or script; close after extracting what you need
- `"startup"` or `"user"` — leave these alone unless the task requires it
-
-**Cleanup tools:**
- `browser_close(target_id=...)` — close one specific tab
- `browser_close_finished()` — close all your agent/popup tabs (safe: leaves startup/user tabs)
- `browser_close_all()` — close everything except the active tab (use only for full reset)
-
-**Multi-tab workflow pattern:**
-1. Open background tabs with `browser_open(url=..., background=true)` to stay on current tab
-2. Process each tab and close it with `browser_close` when done
-3. When the full workflow completes, call `browser_close_finished()` to confirm cleanup
-4. Check `browser_tabs` at any point — it shows `origin` and `age_seconds` per tab
-
-Never accumulate tabs. Treat every tab you open as a resource you must free.
-
-## Login & Auth Walls
- If you see a "Log in" or "Sign up" prompt instead of expected
-  content, report the auth wall immediately — do NOT attempt to log in.
- Check for cookie consent banners and dismiss them if they block content.
-
-## Efficiency
- Minimize tool calls — combine actions where possible.
- When a snapshot result is saved to a spillover file, use
-  `run_command` with grep to extract specific data rather than
-  re-reading the full file.
- Call `set_output` in the same turn as your last browser action
-  when possible — don't waste a turn.
-"""
@@ -0,0 +1,15 @@
+"""Host layer -- how agents are triggered and hosted."""
+
+from framework.host.colony_runtime import (  # noqa: F401
+    ColonyConfig,
+    ColonyRuntime,
+    StreamEventBus,
+    TriggerSpec,
+)
+from framework.host.event_bus import AgentEvent, EventBus, EventType  # noqa: F401
+from framework.host.worker import (  # noqa: F401
+    Worker,
+    WorkerInfo,
+    WorkerResult,
+    WorkerStatus,
+)
@@ -0,0 +1,95 @@
+"""Read/write helpers for per-colony metadata.json.
+
+A colony's metadata.json lives at ``{COLONIES_DIR}/{colony_name}/metadata.json``
+and holds immutable provenance: the queen that created it, the forked
+session id, creation/update timestamps, and the list of workers.
+
+Mutable user-editable tool configuration lives in a sibling
+``tools.json`` sidecar — see :mod:`framework.host.colony_tools_config`
+— so identity and tool gating evolve independently.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+from pathlib import Path
+from typing import Any
+
+from framework.config import COLONIES_DIR
+
+logger = logging.getLogger(__name__)
+
+
+def colony_metadata_path(colony_name: str) -> Path:
+    """Return the on-disk path to a colony's metadata.json."""
+    return COLONIES_DIR / colony_name / "metadata.json"
+
+
+def load_colony_metadata(colony_name: str) -> dict[str, Any]:
+    """Load metadata.json for ``colony_name``.
+
+    Returns an empty dict if the file is missing or malformed — callers
+    are expected to treat missing fields as defaults.
+    """
+    path = colony_metadata_path(colony_name)
+    if not path.exists():
+        return {}
+    try:
+        data = json.loads(path.read_text(encoding="utf-8"))
+    except (json.JSONDecodeError, OSError):
+        logger.warning("Failed to read colony metadata at %s", path)
+        return {}
+    return data if isinstance(data, dict) else {}
+
+
+def update_colony_metadata(colony_name: str, updates: dict[str, Any]) -> dict[str, Any]:
+    """Shallow-merge ``updates`` into metadata.json and persist.
+
+    Returns the full updated dict. Raises ``FileNotFoundError`` if the
+    colony does not exist. Writes atomically via ``os.replace`` to
+    minimize the window where a reader could see a half-written file.
+    """
+    import os
+    import tempfile
+
+    path = colony_metadata_path(colony_name)
+    if not path.parent.exists():
+        raise FileNotFoundError(f"Colony '{colony_name}' not found")
+
+    data = load_colony_metadata(colony_name) if path.exists() else {}
+    for key, value in updates.items():
+        data[key] = value
+
+    path.parent.mkdir(parents=True, exist_ok=True)
+    fd, tmp_path = tempfile.mkstemp(
+        prefix=".metadata.",
+        suffix=".json.tmp",
+        dir=str(path.parent),
+    )
+    try:
+        with os.fdopen(fd, "w", encoding="utf-8") as fh:
+            json.dump(data, fh, indent=2)
+            fh.flush()
+            os.fsync(fh.fileno())
+        os.replace(tmp_path, path)
+    except BaseException:
+        try:
+            os.unlink(tmp_path)
+        except OSError:
+            pass
+        raise
+    return data
+
+
+def list_colony_names() -> list[str]:
+    """Return the names of every colony that has a metadata.json on disk."""
+    if not COLONIES_DIR.is_dir():
+        return []
+    names: list[str] = []
+    for entry in sorted(COLONIES_DIR.iterdir()):
+        if not entry.is_dir():
+            continue
+        if (entry / "metadata.json").exists():
+            names.append(entry.name)
+    return names
@@ -0,0 +1,162 @@
+"""Per-colony tool configuration sidecar (``tools.json``).
+
+Lives at ``~/.hive/colonies/{colony_name}/tools.json`` alongside
+``metadata.json``. Kept separate so provenance (queen_name,
+created_at, workers) stays in metadata while the user-editable tool
+allowlist gets its own file.
+
+Schema::
+
+    {
+      "enabled_mcp_tools": ["read_file", ...] | null,
+      "updated_at": "2026-04-21T12:34:56+00:00"
+    }
+
+- ``null`` / missing file → default "allow every MCP tool".
+- ``[]`` → explicitly disable every MCP tool.
+- ``["foo", "bar"]`` → only those MCP tool names pass the filter.
+
+Atomic writes via ``os.replace`` mirror
+``framework.host.colony_metadata.update_colony_metadata``.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import os
+import tempfile
+from datetime import UTC, datetime
+from pathlib import Path
+from typing import Any
+
+from framework.config import COLONIES_DIR
+
+logger = logging.getLogger(__name__)
+
+
+def tools_config_path(colony_name: str) -> Path:
+    """Return the on-disk path to a colony's ``tools.json``."""
+    return COLONIES_DIR / colony_name / "tools.json"
+
+
+def _metadata_path(colony_name: str) -> Path:
+    return COLONIES_DIR / colony_name / "metadata.json"
+
+
+def _atomic_write_json(path: Path, data: dict[str, Any]) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    fd, tmp = tempfile.mkstemp(
+        prefix=".tools.",
+        suffix=".json.tmp",
+        dir=str(path.parent),
+    )
+    try:
+        with os.fdopen(fd, "w", encoding="utf-8") as fh:
+            json.dump(data, fh, indent=2)
+            fh.flush()
+            os.fsync(fh.fileno())
+        os.replace(tmp, path)
+    except BaseException:
+        try:
+            os.unlink(tmp)
+        except OSError:
+            pass
+        raise
+
+
+def _migrate_from_metadata_if_needed(colony_name: str) -> list[str] | None:
+    """Hoist a legacy ``enabled_mcp_tools`` field out of ``metadata.json``.
+
+    Returns the migrated value (or ``None`` if nothing to migrate). After
+    migration the sidecar exists and ``metadata.json`` no longer contains
+    ``enabled_mcp_tools``. Safe to call repeatedly.
+    """
+    meta_path = _metadata_path(colony_name)
+    if not meta_path.exists():
+        return None
+    try:
+        data = json.loads(meta_path.read_text(encoding="utf-8"))
+    except (json.JSONDecodeError, OSError):
+        logger.warning("Could not read metadata.json during tools migration: %s", colony_name)
+        return None
+    if not isinstance(data, dict) or "enabled_mcp_tools" not in data:
+        return None
+
+    raw = data.pop("enabled_mcp_tools")
+    enabled: list[str] | None
+    if raw is None:
+        enabled = None
+    elif isinstance(raw, list) and all(isinstance(x, str) for x in raw):
+        enabled = raw
+    else:
+        logger.warning(
+            "Legacy enabled_mcp_tools on colony %s had unexpected shape %r; dropping",
+            colony_name,
+            raw,
+        )
+        enabled = None
+
+    # Sidecar first so a partial failure leaves the config recoverable.
+    _atomic_write_json(
+        tools_config_path(colony_name),
+        {
+            "enabled_mcp_tools": enabled,
+            "updated_at": datetime.now(UTC).isoformat(),
+        },
+    )
+    _atomic_write_json(meta_path, data)
+    logger.info(
+        "Migrated enabled_mcp_tools for colony %s from metadata.json to tools.json",
+        colony_name,
+    )
+    return enabled
+
+
+def load_colony_tools_config(colony_name: str) -> list[str] | None:
+    """Return the colony's MCP tool allowlist, or ``None`` for default-allow.
+
+    Order of resolution:
+    1. ``tools.json`` sidecar (authoritative).
+    2. Legacy ``metadata.json`` field (migrated and deleted on first read).
+    3. ``None`` — default "allow every MCP tool".
+    """
+    path = tools_config_path(colony_name)
+    if path.exists():
+        try:
+            data = json.loads(path.read_text(encoding="utf-8"))
+        except (json.JSONDecodeError, OSError):
+            logger.warning("Invalid %s; treating as default-allow", path)
+            return None
+        if not isinstance(data, dict):
+            return None
+        raw = data.get("enabled_mcp_tools")
+        if raw is None:
+            return None
+        if isinstance(raw, list) and all(isinstance(x, str) for x in raw):
+            return raw
+        logger.warning("Unexpected enabled_mcp_tools shape in %s; ignoring", path)
+        return None
+
+    return _migrate_from_metadata_if_needed(colony_name)
+
+
+def update_colony_tools_config(
+    colony_name: str,
+    enabled_mcp_tools: list[str] | None,
+) -> list[str] | None:
+    """Persist a colony's MCP allowlist to ``tools.json``.
+
+    Raises ``FileNotFoundError`` if the colony's directory is missing.
+    """
+    colony_dir = COLONIES_DIR / colony_name
+    if not colony_dir.exists():
+        raise FileNotFoundError(f"Colony directory not found: {colony_name}")
+    _atomic_write_json(
+        tools_config_path(colony_name),
+        {
+            "enabled_mcp_tools": enabled_mcp_tools,
+            "updated_at": datetime.now(UTC).isoformat(),
+        },
+    )
+    return enabled_mcp_tools
@@ -108,14 +108,19 @@ class EventType(StrEnum):
    # Judge decisions (implicit judge in event loop nodes)
    JUDGE_VERDICT = "judge_verdict"

-    # Output tracking
-    OUTPUT_KEY_SET = "output_key_set"
-
-    # Retry / edge tracking
+    # Retry tracking
    NODE_RETRY = "node_retry"
-    EDGE_TRAVERSED = "edge_traversed"

-    # Worker agent lifecycle (event-driven graph execution)
+    # Stream-health observability. Split from NODE_RETRY so the UI can
+    # distinguish "slow TTFT on a huge context" (healthy, just slow) from
+    # "stream went silent mid-generation" (probable stall) from "we nudged
+    # the model to continue" (recovery), which NODE_RETRY used to conflate.
+    STREAM_TTFT_EXCEEDED = "stream_ttft_exceeded"
+    STREAM_INACTIVE = "stream_inactive"
+    STREAM_NUDGE_SENT = "stream_nudge_sent"
+    TOOL_CALL_REPLAY_DETECTED = "tool_call_replay_detected"
+
+    # Worker agent lifecycle
    WORKER_COMPLETED = "worker_completed"
    WORKER_FAILED = "worker_failed"

@@ -135,21 +140,19 @@ class EventType(StrEnum):
    # Execution resurrection (auto-restart on non-fatal failure)
    EXECUTION_RESURRECTED = "execution_resurrected"

-    # Graph lifecycle (session manager → frontend)
-    WORKER_GRAPH_LOADED = "worker_graph_loaded"
+    # Colony lifecycle (session manager → frontend)
+    WORKER_COLONY_LOADED = "worker_colony_loaded"
+    # Queen create_colony tool finished forking; carries colony_name +
+    # path so the frontend can render a system message linking to the
+    # new colony page at /colony/{colony_name}.
+    COLONY_CREATED = "colony_created"
    CREDENTIALS_REQUIRED = "credentials_required"

-    # Draft graph (planning phase — lightweight graph preview)
-    DRAFT_GRAPH_UPDATED = "draft_graph_updated"
-
-    # Flowchart map updated (after reconciliation with runtime graph)
-    FLOWCHART_MAP_UPDATED = "flowchart_map_updated"
-
-    # Queen phase changes (building <-> staging <-> running)
+    # Queen phase changes (working <-> reviewing)
    QUEEN_PHASE_CHANGED = "queen_phase_changed"

-    # Queen thinking hook — persona selected for the current building session
-    QUEEN_PERSONA_SELECTED = "queen_persona_selected"
+    # Queen identity — which queen profile was selected for this session
+    QUEEN_IDENTITY_SELECTED = "queen_identity_selected"

    # Subagent reports (one-way progress updates from sub-agents)
    SUBAGENT_REPORT = "subagent_report"
@@ -174,7 +177,7 @@ class AgentEvent:
    data: dict[str, Any] = field(default_factory=dict)
    timestamp: datetime = field(default_factory=datetime.now)
    correlation_id: str | None = None  # For tracking related events
-    graph_id: str | None = None  # Which graph emitted this event (multi-graph sessions)
+    colony_id: str | None = None  # Which colony emitted this event
    run_id: str | None = None  # Unique ID per trigger() invocation — used for run dividers

    def to_dict(self) -> dict:
@@ -187,7 +190,7 @@ class AgentEvent:
            "data": self.data,
            "timestamp": self.timestamp.isoformat(),
            "correlation_id": self.correlation_id,
-            "graph_id": self.graph_id,
+            "colony_id": self.colony_id,
        }
        if self.run_id is not None:
            d["run_id"] = self.run_id
@@ -208,7 +211,7 @@ class Subscription:
    filter_stream: str | None = None  # Only receive events from this stream
    filter_node: str | None = None  # Only receive events from this node
    filter_execution: str | None = None  # Only receive events from this execution
-    filter_graph: str | None = None  # Only receive events from this graph
+    filter_colony: str | None = None  # Only receive events from this colony


 class EventBus:
@@ -390,7 +393,7 @@ class EventBus:
        filter_stream: str | None = None,
        filter_node: str | None = None,
        filter_execution: str | None = None,
-        filter_graph: str | None = None,
+        filter_colony: str | None = None,
    ) -> str:
        """
        Subscribe to events.
@@ -401,7 +404,7 @@ class EventBus:
            filter_stream: Only receive events from this stream
            filter_node: Only receive events from this node
            filter_execution: Only receive events from this execution
-            filter_graph: Only receive events from this graph
+            filter_colony: Only receive events from this colony

        Returns:
            Subscription ID (use to unsubscribe)
@@ -416,7 +419,7 @@ class EventBus:
            filter_stream=filter_stream,
            filter_node=filter_node,
            filter_execution=filter_execution,
-            filter_graph=filter_graph,
+            filter_colony=filter_colony,
        )

        self._subscriptions[sub_id] = subscription
@@ -452,11 +455,7 @@ class EventBus:
        # iteration values.  Without this, live SSE would use raw iterations
        # while events.jsonl would use offset iterations, causing ID collisions
        # on the frontend when replaying after cold resume.
-        if (
-            self._session_log_iteration_offset
-            and isinstance(event.data, dict)
-            and "iteration" in event.data
-        ):
+        if self._session_log_iteration_offset and isinstance(event.data, dict) and "iteration" in event.data:
            offset = self._session_log_iteration_offset
            event.data = {**event.data, "iteration": event.data["iteration"] + offset}

@@ -518,23 +517,41 @@ class EventBus:
        if subscription.filter_execution and subscription.filter_execution != event.execution_id:
            return False

-        # Check graph filter
-        if subscription.filter_graph and subscription.filter_graph != event.graph_id:
+        # Check colony filter
+        if subscription.filter_colony and subscription.filter_colony != event.colony_id:
            return False

        return True

+    # Per-handler wall-clock timeout. A subscriber that deadlocks or
+    # blocks on slow I/O would otherwise freeze the publisher (and via
+    # ``await publish(...)`` any coroutine that emits events) indefinitely.
+    # 15 s is generous for legitimate handlers and cheap to tune later.
+    _HANDLER_TIMEOUT_SECONDS: float = 15.0
+
    async def _execute_handlers(
        self,
        event: AgentEvent,
        handlers: list[EventHandler],
    ) -> None:
-        """Execute handlers concurrently with rate limiting."""
+        """Execute handlers concurrently with rate limiting + hard timeout."""

        async def run_handler(handler: EventHandler) -> None:
            async with self._semaphore:
                try:
-                    await handler(event)
+                    await asyncio.wait_for(
+                        handler(event),
+                        timeout=self._HANDLER_TIMEOUT_SECONDS,
+                    )
+                except TimeoutError:
+                    handler_name = getattr(handler, "__qualname__", repr(handler))
+                    logger.error(
+                        "EventBus handler %s exceeded %.0fs on event %s — dropping; "
+                        "fix the handler or the publisher will stall",
+                        handler_name,
+                        self._HANDLER_TIMEOUT_SECONDS,
+                        getattr(event.type, "name", event.type),
+                    )
                except Exception:
                    logger.exception(f"Handler error for {event.type}")

@@ -792,16 +809,28 @@ class EventBus:
        input_tokens: int,
        output_tokens: int,
        cached_tokens: int = 0,
+        cache_creation_tokens: int = 0,
+        cost_usd: float = 0.0,
        execution_id: str | None = None,
        iteration: int | None = None,
    ) -> None:
-        """Emit LLM turn completion with stop reason and model metadata."""
+        """Emit LLM turn completion with stop reason and model metadata.
+
+        ``cached_tokens`` and ``cache_creation_tokens`` are subsets of
+        ``input_tokens`` (already inside provider ``prompt_tokens``).
+        Subscribers should display them, not add them to a total.
+
+        ``cost_usd`` is the USD cost for this turn when known (Anthropic,
+        OpenAI, OpenRouter). 0.0 means unreported (not free).
+        """
        data: dict = {
            "stop_reason": stop_reason,
            "model": model,
            "input_tokens": input_tokens,
            "output_tokens": output_tokens,
            "cached_tokens": cached_tokens,
+            "cache_creation_tokens": cache_creation_tokens,
+            "cost_usd": cost_usd,
        }
        if iteration is not None:
            data["iteration"] = iteration
@@ -897,24 +926,22 @@ class EventBus:
        self,
        stream_id: str,
        node_id: str,
-        prompt: str = "",
        execution_id: str | None = None,
-        options: list[str] | None = None,
        questions: list[dict] | None = None,
    ) -> None:
        """Emit a user-input request for interactive queen turns.

        Args:
-            options: Optional predefined choices for the user (1-3 items).
-                     The frontend appends an "Other" free-text option
-                     automatically.
-            questions: Optional list of question dicts for multi-question
-                       batches (from ask_user_multiple). Each dict has id,
-                       prompt, and optional options.
+            questions: Optional list of question dicts from ``ask_user``.
+                Each dict has ``id``, ``prompt``, and optional ``options``
+                (2-3 predefined choices). The frontend renders the
+                QuestionWidget for a single-entry list and the
+                MultiQuestionWidget for 2+ entries. Free-text asks (no
+                options) stream the prompt separately as a chat message;
+                auto-block turns have no questions at all and fall back
+                to the normal text input.
        """
-        data: dict[str, Any] = {"prompt": prompt}
-        if options:
-            data["options"] = options
+        data: dict[str, Any] = {}
        if questions:
            data["questions"] = questions
        await self.publish(
@@ -1029,24 +1056,6 @@ class EventBus:
            )
        )

-    async def emit_output_key_set(
-        self,
-        stream_id: str,
-        node_id: str,
-        key: str,
-        execution_id: str | None = None,
-    ) -> None:
-        """Emit output key set event."""
-        await self.publish(
-            AgentEvent(
-                type=EventType.OUTPUT_KEY_SET,
-                stream_id=stream_id,
-                node_id=node_id,
-                execution_id=execution_id,
-                data={"key": key},
-            )
-        )
-
    async def emit_node_retry(
        self,
        stream_id: str,
@@ -1071,25 +1080,90 @@ class EventBus:
            )
        )

-    async def emit_edge_traversed(
+    async def emit_stream_ttft_exceeded(
        self,
        stream_id: str,
-        source_node: str,
-        target_node: str,
-        edge_condition: str = "",
+        node_id: str,
+        ttft_seconds: float,
+        limit_seconds: float,
        execution_id: str | None = None,
    ) -> None:
-        """Emit edge traversed event."""
+        """Emit when a stream stayed silent past the TTFT budget (no first event)."""
        await self.publish(
            AgentEvent(
-                type=EventType.EDGE_TRAVERSED,
+                type=EventType.STREAM_TTFT_EXCEEDED,
                stream_id=stream_id,
-                node_id=source_node,
+                node_id=node_id,
                execution_id=execution_id,
                data={
-                    "source_node": source_node,
-                    "target_node": target_node,
-                    "edge_condition": edge_condition,
+                    "ttft_seconds": ttft_seconds,
+                    "limit_seconds": limit_seconds,
+                },
+            )
+        )
+
+    async def emit_stream_inactive(
+        self,
+        stream_id: str,
+        node_id: str,
+        idle_seconds: float,
+        limit_seconds: float,
+        execution_id: str | None = None,
+    ) -> None:
+        """Emit when a stream that had produced events went silent past budget."""
+        await self.publish(
+            AgentEvent(
+                type=EventType.STREAM_INACTIVE,
+                stream_id=stream_id,
+                node_id=node_id,
+                execution_id=execution_id,
+                data={
+                    "idle_seconds": idle_seconds,
+                    "limit_seconds": limit_seconds,
+                },
+            )
+        )
+
+    async def emit_stream_nudge_sent(
+        self,
+        stream_id: str,
+        node_id: str,
+        reason: str,
+        nudge_count: int,
+        execution_id: str | None = None,
+    ) -> None:
+        """Emit when the continue-nudge was injected (recovery, not retry)."""
+        await self.publish(
+            AgentEvent(
+                type=EventType.STREAM_NUDGE_SENT,
+                stream_id=stream_id,
+                node_id=node_id,
+                execution_id=execution_id,
+                data={
+                    "reason": reason,
+                    "nudge_count": nudge_count,
+                },
+            )
+        )
+
+    async def emit_tool_call_replay_detected(
+        self,
+        stream_id: str,
+        node_id: str,
+        tool_name: str,
+        prior_seq: int,
+        execution_id: str | None = None,
+    ) -> None:
+        """Emit when the model is about to re-execute a prior successful call."""
+        await self.publish(
+            AgentEvent(
+                type=EventType.TOOL_CALL_REPLAY_DETECTED,
+                stream_id=stream_id,
+                node_id=node_id,
+                execution_id=execution_id,
+                data={
+                    "tool_name": tool_name,
+                    "prior_seq": prior_seq,
                },
            )
        )
@@ -1208,15 +1282,25 @@ class EventBus:
        reason: str = "",
        context: str = "",
        execution_id: str | None = None,
+        request_id: str | None = None,
    ) -> None:
-        """Emit escalation requested event (agent wants queen)."""
+        """Emit escalation requested event (agent wants queen).
+
+        ``request_id`` is a caller-supplied handle used by the queen to
+        address its reply back to the specific escalation. When omitted the
+        event still fires but the queen cannot route a targeted reply.
+        """
        await self.publish(
            AgentEvent(
                type=EventType.ESCALATION_REQUESTED,
                stream_id=stream_id,
                node_id=node_id,
                execution_id=execution_id,
-                data={"reason": reason, "context": context},
+                data={
+                    "request_id": request_id,
+                    "reason": reason,
+                    "context": context,
+                },
            )
        )

@@ -1297,7 +1381,7 @@ class EventBus:
        stream_id: str | None = None,
        node_id: str | None = None,
        execution_id: str | None = None,
-        graph_id: str | None = None,
+        colony_id: str | None = None,
        timeout: float | None = None,
    ) -> AgentEvent | None:
        """
@@ -1308,7 +1392,7 @@ class EventBus:
            stream_id: Filter by stream
            node_id: Filter by node
            execution_id: Filter by execution
-            graph_id: Filter by graph
+            colony_id: Filter by colony
            timeout: Maximum time to wait (seconds)

        Returns:
@@ -1329,7 +1413,7 @@ class EventBus:
            filter_stream=stream_id,
            filter_node=node_id,
            filter_execution=execution_id,
-            filter_graph=graph_id,
+            filter_colony=colony_id,
        )

        try:
@@ -16,20 +16,20 @@ from collections import OrderedDict
 from collections.abc import Callable
 from dataclasses import dataclass, field
 from datetime import datetime
-from typing import TYPE_CHECKING, Any
+from typing import TYPE_CHECKING, Any, Literal

-from framework.graph.checkpoint_config import CheckpointConfig
-from framework.graph.executor import ExecutionResult, GraphExecutor
-from framework.runtime.event_bus import EventBus
-from framework.runtime.shared_state import IsolationLevel, SharedBufferManager
-from framework.runtime.stream_runtime import StreamRuntime, StreamRuntimeAdapter
+from framework.host.event_bus import EventBus
+from framework.host.shared_state import IsolationLevel, SharedBufferManager
+from framework.host.stream_runtime import StreamDecisionTracker, StreamRuntimeAdapter
+from framework.orchestrator.checkpoint_config import CheckpointConfig
+from framework.orchestrator.orchestrator import ExecutionResult, Orchestrator

 if TYPE_CHECKING:
-    from framework.graph.edge import GraphSpec
-    from framework.graph.goal import Goal
+    from framework.host.event_bus import AgentEvent
+    from framework.host.outcome_aggregator import OutcomeAggregator
    from framework.llm.provider import LLMProvider, Tool
-    from framework.runtime.event_bus import AgentEvent
-    from framework.runtime.outcome_aggregator import OutcomeAggregator
+    from framework.orchestrator.edge import GraphSpec
+    from framework.orchestrator.goal import Goal
    from framework.storage.concurrent import ConcurrentStorage
    from framework.storage.session_store import SessionStore

@@ -48,6 +48,8 @@ class ExecutionAlreadyRunningError(RuntimeError):

 logger = logging.getLogger(__name__)

+CancelExecutionResult = Literal["cancelled", "cancelling", "not_found"]
+

 class GraphScopedEventBus(EventBus):
    """Proxy that stamps ``graph_id`` on every published event.
@@ -130,10 +132,10 @@ class ExecutionContext:
    run_id: str | None = None  # Unique ID per trigger() invocation
    started_at: datetime = field(default_factory=datetime.now)
    completed_at: datetime | None = None
-    status: str = "pending"  # pending, running, completed, failed, paused
+    status: str = "pending"  # pending, running, cancelling, completed, failed, paused, cancelled


-class ExecutionStream:
+class ExecutionManager:
    """
    Manages concurrent executions for a single entry point.

@@ -172,7 +174,7 @@ class ExecutionStream:
        goal: "Goal",
        state_manager: SharedBufferManager,
        storage: "ConcurrentStorage",
-        outcome_aggregator: "OutcomeAggregator",
+        outcome_aggregator: "OutcomeAggregator | None" = None,
        event_bus: "EventBus | None" = None,
        llm: "LLMProvider | None" = None,
        tools: list["Tool"] | None = None,
@@ -192,10 +194,6 @@ class ExecutionStream:
        context_warn_ratio: float | None = None,
        batch_init_nudge: str | None = None,
        dynamic_memory_provider_factory: Callable[[str], Callable[[], str] | None] | None = None,
-        colony_memory_dir: Any = None,
-        colony_worker_sessions_dir: Any = None,
-        colony_recall_cache: dict[str, str] | None = None,
-        colony_reflect_llm: Any = None,
    ):
        """
        Initialize execution stream.
@@ -251,10 +249,6 @@ class ExecutionStream:
        self._context_warn_ratio: float | None = context_warn_ratio
        self._batch_init_nudge: str | None = batch_init_nudge
        self._dynamic_memory_provider_factory = dynamic_memory_provider_factory
-        self._colony_memory_dir = colony_memory_dir
-        self._colony_worker_sessions_dir = colony_worker_sessions_dir
-        self._colony_recall_cache = colony_recall_cache
-        self._colony_reflect_llm = colony_reflect_llm

        _es_logger = logging.getLogger(__name__)
        if protocols_prompt:
@@ -270,16 +264,15 @@ class ExecutionStream:
            )

        # Create stream-scoped runtime
-        self._runtime = StreamRuntime(
+        self._runtime = StreamDecisionTracker(
            stream_id=stream_id,
            storage=storage,
-            outcome_aggregator=outcome_aggregator,
        )

        # Execution tracking
        self._active_executions: dict[str, ExecutionContext] = {}
        self._execution_tasks: dict[str, asyncio.Task] = {}
-        self._active_executors: dict[str, GraphExecutor] = {}
+        self._active_executors: dict[str, Orchestrator] = {}
        self._cancel_reasons: dict[str, str] = {}
        self._execution_results: OrderedDict[str, ExecutionResult] = OrderedDict()
        self._execution_result_times: dict[str, float] = {}
@@ -309,7 +302,7 @@ class ExecutionStream:

        # Emit stream started event
        if self._scoped_event_bus:
-            from framework.runtime.event_bus import AgentEvent, EventType
+            from framework.host.event_bus import AgentEvent, EventType

            await self._scoped_event_bus.publish(
                AgentEvent(
@@ -324,6 +317,22 @@ class ExecutionStream:
        """Return IDs of all currently active executions."""
        return list(self._active_executions.keys())

+    def _get_blocking_execution_ids_locked(self) -> list[str]:
+        """Return executions that still block a replacement from starting.
+
+        An execution continues to block replacement until its task has
+        terminated and the task's final cleanup has removed its bookkeeping.
+        This is intentional: a timed-out cancellation does not mean the old
+        task is harmless. If it is still alive, it can still write shared
+        session state, so letting a replacement start would guarantee
+        overlapping mutations on the same session.
+        """
+        blocking_ids: list[str] = list(self._active_executions.keys())
+        for execution_id, task in self._execution_tasks.items():
+            if not task.done() and execution_id not in self._active_executions:
+                blocking_ids.append(execution_id)
+        return blocking_ids
+
    @property
    def agent_idle_seconds(self) -> float:
        """Seconds since the last agent activity (LLM call, tool call, node transition).
@@ -405,15 +414,22 @@ class ExecutionStream:

    async def stop(self) -> None:
        """Stop the execution stream and cancel active executions."""
-        if not self._running:
-            return
+        async with self._lock:
+            if not self._running:
+                return

-        self._running = False
+            self._running = False

-        # Cancel all active executions
-        tasks_to_wait = []
-        for _, task in self._execution_tasks.items():
-            if not task.done():
+            # Cancel all active executions, but keep bookkeeping until each
+            # task reaches its own cleanup path.
+            tasks_to_wait: list[asyncio.Task] = []
+            for execution_id, task in self._execution_tasks.items():
+                if task.done():
+                    continue
+                ctx = self._active_executions.get(execution_id)
+                if ctx is not None:
+                    ctx.status = "cancelling"
+                self._cancel_reasons.setdefault(execution_id, "Execution cancelled")
                task.cancel()
                tasks_to_wait.append(task)

@@ -427,14 +443,11 @@ class ExecutionStream:
                    len(pending),
                )

-        self._execution_tasks.clear()
-        self._active_executions.clear()
-
        logger.info(f"ExecutionStream '{self.stream_id}' stopped")

        # Emit stream stopped event
        if self._scoped_event_bus:
-            from framework.runtime.event_bus import AgentEvent, EventType
+            from framework.host.event_bus import AgentEvent, EventType

            await self._scoped_event_bus.publish(
                AgentEvent(
@@ -461,9 +474,7 @@ class ExecutionStream:
        for executor in self._active_executors.values():
            node = executor.node_registry.get(node_id)
            if node is not None and hasattr(node, "inject_event"):
-                await node.inject_event(
-                    content, is_client_input=is_client_input, image_content=image_content
-                )
+                await node.inject_event(content, is_client_input=is_client_input, image_content=image_content)
                return True
        return False

@@ -580,12 +591,16 @@ class ExecutionStream:
        )

        async with self._lock:
+            if not self._running:
+                raise RuntimeError(f"ExecutionStream '{self.stream_id}' is not running")
+
+            blocking_ids = self._get_blocking_execution_ids_locked()
+            if blocking_ids:
+                raise ExecutionAlreadyRunningError(self.stream_id, blocking_ids)
+
            self._active_executions[execution_id] = ctx
            self._completion_events[execution_id] = asyncio.Event()
-
-        # Start execution task
-        task = asyncio.create_task(self._run_execution(ctx))
-        self._execution_tasks[execution_id] = task
+            self._execution_tasks[execution_id] = asyncio.create_task(self._run_execution(ctx))

        logger.debug(f"Queued execution {execution_id} for stream {self.stream_id}")
        return execution_id
@@ -676,11 +691,9 @@ class ExecutionStream:
                # Create per-execution runtime logger
                runtime_logger = None
                if self._runtime_log_store:
-                    from framework.runtime.runtime_logger import RuntimeLogger
+                    from framework.tracker.runtime_logger import RuntimeLogger

-                    runtime_logger = RuntimeLogger(
-                        store=self._runtime_log_store, agent_id=self.graph.id
-                    )
+                    runtime_logger = RuntimeLogger(store=self._runtime_log_store, agent_id=self.graph.id)

                # Derive storage from session_store (graph-specific for secondary
                # graphs) so that all files — conversations, state, checkpoints,
@@ -705,12 +718,7 @@ class ExecutionStream:
                # forward so the next attempt resumes at the failed node.
                while True:
                    # Create executor for this execution.
-                    # Each execution gets its own storage under sessions/{exec_id}/
-                    # so conversations, spillover, and data files are all scoped
-                    # to this execution.  The executor sets data_dir via execution
-                    # context (contextvars) so data tools and spillover share the
-                    # same session-scoped directory.
-                    executor = GraphExecutor(
+                    executor = Orchestrator(
                        runtime=runtime_adapter,
                        llm=self._llm,
                        tools=self._tools,
@@ -735,10 +743,6 @@ class ExecutionStream:
                            if self._dynamic_memory_provider_factory is not None
                            else None
                        ),
-                        colony_memory_dir=self._colony_memory_dir,
-                        colony_worker_sessions_dir=self._colony_worker_sessions_dir,
-                        colony_recall_cache=self._colony_recall_cache,
-                        colony_reflect_llm=self._colony_reflect_llm,
                    )
                    # Track executor so inject_input() can reach EventLoopNode instances
                    self._active_executors[execution_id] = executor
@@ -775,7 +779,7 @@ class ExecutionStream:

                        # Emit resurrection event
                        if self._scoped_event_bus:
-                            from framework.runtime.event_bus import AgentEvent, EventType
+                            from framework.host.event_bus import AgentEvent, EventType

                            await self._scoped_event_bus.publish(
                                AgentEvent(
@@ -905,9 +909,7 @@ class ExecutionStream:
                    if has_result and result.paused_at:
                        await self._write_session_state(execution_id, ctx, result=result)
                    else:
-                        await self._write_session_state(
-                            execution_id, ctx, error="Execution cancelled"
-                        )
+                        await self._write_session_state(execution_id, ctx, error="Execution cancelled")

                # Emit SSE event so the frontend knows the execution stopped.
                # The executor does NOT emit on CancelledError, so there is no
@@ -1131,7 +1133,7 @@ class ExecutionStream:
        Each stream only executes from its own entry_node, but the full
        graph must validate with all entry points accounted for.
        """
-        from framework.graph.edge import GraphSpec
+        from framework.orchestrator.edge import GraphSpec

        # Merge entry points: this stream's entry + original graph's primary
        # entry + any other entry points. This ensures all nodes are
@@ -1207,7 +1209,7 @@ class ExecutionStream:
        """Get execution context."""
        return self._active_executions.get(execution_id)

-    async def cancel_execution(self, execution_id: str, *, reason: str | None = None) -> bool:
+    async def cancel_execution(self, execution_id: str, *, reason: str | None = None) -> CancelExecutionResult:
        """
        Cancel a running execution.

@@ -1218,34 +1220,38 @@ class ExecutionStream:
                provided, defaults to "Execution cancelled".

        Returns:
-            True if cancelled, False if not found
+            "cancelled" if the task fully exited within the grace period,
+            "cancelling" if cancellation was requested but the task is still
+            shutting down, or "not_found" if no active task exists.
        """
-        task = self._execution_tasks.get(execution_id)
-        if task and not task.done():
+        async with self._lock:
+            task = self._execution_tasks.get(execution_id)
+            if task is None or task.done():
+                return "not_found"
+
            # Store the reason so the CancelledError handler can use it
            # when emitting the pause/fail event.
            self._cancel_reasons[execution_id] = reason or "Execution cancelled"
+            ctx = self._active_executions.get(execution_id)
+            if ctx is not None:
+                ctx.status = "cancelling"
            task.cancel()
-            # Wait briefly for the task to finish. Don't block indefinitely —
-            # the task may be stuck in a long LLM API call that doesn't
-            # respond to cancellation quickly.
-            done, _ = await asyncio.wait({task}, timeout=5.0)
-            if not done:
-                # Task didn't finish within timeout — clean up bookkeeping now
-                # so the session doesn't think it still has running executions.
-                # The task will continue winding down in the background and its
-                # finally block will harmlessly pop already-removed keys.
-                logger.warning(
-                    "Execution %s did not finish within cancel timeout; "
-                    "force-cleaning bookkeeping",
-                    execution_id,
-                )
-                async with self._lock:
-                    self._active_executions.pop(execution_id, None)
-                    self._execution_tasks.pop(execution_id, None)
-                self._active_executors.pop(execution_id, None)
-            return True
-        return False
+
+        # Wait briefly for the task to finish. Don't block indefinitely —
+        # the task may be stuck in a long LLM API call that doesn't
+        # respond to cancellation quickly.
+        done, _ = await asyncio.wait({task}, timeout=5.0)
+        if not done:
+            # Keep bookkeeping in place until the task's own finally block runs.
+            # We intentionally do not add deferred cleanup keyed by execution_id
+            # here because resumed executions reuse the same id; a delayed pop
+            # could otherwise delete bookkeeping that belongs to the new run.
+            logger.warning(
+                "Execution %s did not finish within cancel timeout; leaving bookkeeping in place until task exit",
+                execution_id,
+            )
+            return "cancelling"
+        return "cancelled"

    # === STATS AND MONITORING ===

@@ -0,0 +1,9 @@
+"""State isolation level enum."""
+
+from enum import StrEnum
+
+
+class IsolationLevel(StrEnum):
+    ISOLATED = "isolated"
+    SHARED = "shared"
+    SYNCHRONIZED = "synchronized"
@@ -0,0 +1,21 @@
+"""Stub — outcome aggregator removed in colony refactor."""
+
+from framework.schemas.goal import Goal
+
+
+class OutcomeAggregator:
+    def __init__(self, goal: Goal, event_bus=None):
+        self._goal = goal
+        self._event_bus = event_bus
+
+    def record_decision(self, **kwargs):
+        pass
+
+    def record_outcome(self, **kwargs):
+        pass
+
+    def evaluate_goal_progress(self):
+        return {"progress": 0.0, "criteria_status": {}}
+
+    def get_stats(self):
+        return {"total_decisions": 0, "total_outcomes": 0}
--- a/Show More
+++ b/Show More