Merge pull request #262 from AgentWorkforce/fix/channel-message-delivery

Fix: Channel message delivery for agents in channels

Author: khaliqgant

.trajectories/completed/2026-01/traj_jl5pp8rdu8up.json

@@ -0,0 +1,61 @@
+{
+  "id": "traj_jl5pp8rdu8up",
+  "version": 1,
+  "task": {
+    "title": "Fix spawn timing race condition"
+  },
+  "status": "completed",
+  "startedAt": "2026-01-21T22:10:45.793Z",
+  "agents": [
+    {
+      "name": "default",
+      "role": "lead",
+      "joinedAt": "2026-01-21T22:29:43.316Z"
+    }
+  ],
+  "chapters": [
+    {
+      "id": "chap_joy3ffeznif4",
+      "title": "Work",
+      "agentName": "default",
+      "startedAt": "2026-01-21T22:29:43.316Z",
+      "events": [
+        {
+          "ts": 1769034583318,
+          "type": "decision",
+          "content": "Synced spawn registration by requiring both connected-agents.json and agents.json freshness before sending spawn tasks: Synced spawn registration by requiring both connected-agents.json and agents.json freshness before sending spawn tasks",
+          "raw": {
+            "question": "Synced spawn registration by requiring both connected-agents.json and agents.json freshness before sending spawn tasks",
+            "chosen": "Synced spawn registration by requiring both connected-agents.json and agents.json freshness before sending spawn tasks",
+            "alternatives": [],
+            "reasoning": "Avoids mismatched online checks and prevents sending before registry is ready"
+          },
+          "significance": "high"
+        },
+        {
+          "ts": 1769034594756,
+          "type": "decision",
+          "content": "Pass spawning callbacks to orchestrator and CLI spawners to keep router queueing aligned: Pass spawning callbacks to orchestrator and CLI spawners to keep router queueing aligned",
+          "raw": {
+            "question": "Pass spawning callbacks to orchestrator and CLI spawners to keep router queueing aligned",
+            "chosen": "Pass spawning callbacks to orchestrator and CLI spawners to keep router queueing aligned",
+            "alternatives": [],
+            "reasoning": "Ensures messages to spawning agents are queued consistently across entrypoints"
+          },
+          "significance": "high"
+        }
+      ],
+      "endedAt": "2026-01-21T22:30:11.795Z"
+    }
+  ],
+  "commits": [],
+  "filesChanged": [],
+  "projectId": "/data/repos/relay",
+  "tags": [],
+  "completedAt": "2026-01-21T22:30:11.795Z",
+  "retrospective": {
+    "summary": "Hardened spawn registration checks against registry timing gaps and added freshness tests",
+    "approach": "Standard approach",
+    "confidence": 0.78
+  }
+}

.trajectories/completed/2026-01/traj_jl5pp8rdu8up.md

@@ -0,0 +1,36 @@
+# Trajectory: Fix spawn timing race condition
+
+> **Status:** ✅ Completed
+> **Confidence:** 78%
+> **Started:** January 21, 2026 at 10:10 PM
+> **Completed:** January 21, 2026 at 10:30 PM
+
+---
+
+## Summary
+
+Hardened spawn registration checks against registry timing gaps and added freshness tests
+
+**Approach:** Standard approach
+
+---
+
+## Key Decisions
+
+### Synced spawn registration by requiring both connected-agents.json and agents.json freshness before sending spawn tasks
+- **Chose:** Synced spawn registration by requiring both connected-agents.json and agents.json freshness before sending spawn tasks
+- **Reasoning:** Avoids mismatched online checks and prevents sending before registry is ready
+
+### Pass spawning callbacks to orchestrator and CLI spawners to keep router queueing aligned
+- **Chose:** Pass spawning callbacks to orchestrator and CLI spawners to keep router queueing aligned
+- **Reasoning:** Ensures messages to spawning agents are queued consistently across entrypoints
+
+---
+
+## Chapters
+
+### 1. Work
+*Agent: default*
+
+- Synced spawn registration by requiring both connected-agents.json and agents.json freshness before sending spawn tasks: Synced spawn registration by requiring both connected-agents.json and agents.json freshness before sending spawn tasks
+- Pass spawning callbacks to orchestrator and CLI spawners to keep router queueing aligned: Pass spawning callbacks to orchestrator and CLI spawners to keep router queueing aligned

.trajectories/index.json

@@ -1,6 +1,6 @@
 {
   "version": 1,
-  "lastUpdated": "2026-01-21T19:08:36.280Z",
+  "lastUpdated": "2026-01-21T22:30:12.356Z",
   "trajectories": {
     "traj_ozd98si6a7ns": {
       "title": "Fix thinking indicator showing on all messages",
@@ -876,6 +876,13 @@
       "startedAt": "2026-01-21T17:36:44.996Z",
       "completedAt": "2026-01-21T19:08:35.954Z",
       "path": "/data/repos/relay/.trajectories/completed/2026-01/traj_mvqthi1oitfo.json"
+    },
+    "traj_jl5pp8rdu8up": {
+      "title": "Fix spawn timing race condition",
+      "status": "completed",
+      "startedAt": "2026-01-21T22:10:45.793Z",
+      "completedAt": "2026-01-21T22:30:11.795Z",
+      "path": "/data/repos/relay/.trajectories/completed/2026-01/traj_jl5pp8rdu8up.json"
     }
   }
 }

src/bridge/spawner.test.ts

@@ -349,6 +349,62 @@ describe('AgentSpawner', () => {
     const output = spawner.getWorkerOutput('Unknown');
     expect(output).toBeNull();
   });
+
+  describe('isAgentRegistered', () => {
+    it('returns true when connected and registry are fresh', () => {
+      const now = new Date().toISOString();
+      readFileSyncMock.mockImplementation((filePath: string) => {
+        if (filePath.includes('connected-agents.json')) {
+          return JSON.stringify({ agents: ['Dev1'], users: [], updatedAt: Date.now() });
+        }
+        if (filePath.includes('agents.json')) {
+          return JSON.stringify({ agents: [{ name: 'Dev1', lastSeen: now }] });
+        }
+        return '';
+      });
+
+      const spawner = new AgentSpawner(projectRoot);
+      const registered = (spawner as any).isAgentRegistered('Dev1');
+
+      expect(registered).toBe(true);
+    });
+
+    it('returns false when connected list is stale', () => {
+      const now = new Date().toISOString();
+      readFileSyncMock.mockImplementation((filePath: string) => {
+        if (filePath.includes('connected-agents.json')) {
+          return JSON.stringify({ agents: ['Dev1'], users: [], updatedAt: Date.now() - 60_000 });
+        }
+        if (filePath.includes('agents.json')) {
+          return JSON.stringify({ agents: [{ name: 'Dev1', lastSeen: now }] });
+        }
+        return '';
+      });
+
+      const spawner = new AgentSpawner(projectRoot);
+      const registered = (spawner as any).isAgentRegistered('Dev1');
+
+      expect(registered).toBe(false);
+    });
+
+    it('returns false when registry is stale', () => {
+      const old = new Date(Date.now() - 60_000).toISOString();
+      readFileSyncMock.mockImplementation((filePath: string) => {
+        if (filePath.includes('connected-agents.json')) {
+          return JSON.stringify({ agents: ['Dev1'], users: [], updatedAt: Date.now() });
+        }
+        if (filePath.includes('agents.json')) {
+          return JSON.stringify({ agents: [{ name: 'Dev1', lastSeen: old }] });
+        }
+        return '';
+      });
+
+      const spawner = new AgentSpawner(projectRoot);
+      const registered = (spawner as any).isAgentRegistered('Dev1');
+
+      expect(registered).toBe(false);
+    });
+  });
 });
 
 describe('readWorkersMetadata', () => {

src/bridge/spawner.ts

@@ -228,8 +228,10 @@ export interface AgentSpawnerOptions {
 }
 
 export class AgentSpawner {
+  private static readonly ONLINE_THRESHOLD_MS = 30_000;
   private activeWorkers: Map<string, ActiveWorker> = new Map();
   private agentsPath: string;
+  private registryPath: string;
   private projectRoot: string;
   private socketPath?: string;
   private logsDir: string;
@@ -255,6 +257,7 @@ export class AgentSpawner {
     // Use connected-agents.json (live socket connections) instead of agents.json (historical registry)
     // This ensures spawned agents have actual daemon connections for channel message delivery
     this.agentsPath = path.join(paths.teamDir, 'connected-agents.json');
+    this.registryPath = path.join(paths.teamDir, 'agents.json');
     this.socketPath = paths.socketPath;
     this.logsDir = path.join(paths.teamDir, 'worker-logs');
     this.workersPath = path.join(paths.teamDir, 'workers.json');
@@ -1055,7 +1058,7 @@ export class AgentSpawner {
   }
 
   /**
-   * Wait for an agent to appear in the registry (agents.json)
+   * Wait for an agent to appear in the connected list and registry (connected-agents.json + agents.json).
    */
   private async waitForAgentRegistration(
     name: string,
@@ -1076,6 +1079,10 @@ export class AgentSpawner {
   }
 
   private isAgentRegistered(name: string): boolean {
+    return this.isAgentConnected(name) && this.isAgentRecentlySeen(name);
+  }
+
+  private isAgentConnected(name: string): boolean {
     if (!this.agentsPath) return false;
     if (!fs.existsSync(this.agentsPath)) return false;
 
@@ -1084,6 +1091,10 @@ export class AgentSpawner {
       // connected-agents.json format: { agents: string[], users: string[], updatedAt: number }
       // agents is a string array of connected agent names (not objects)
       const agents: string[] = Array.isArray(raw?.agents) ? raw.agents : [];
+      const updatedAt = typeof raw?.updatedAt === 'number' ? raw.updatedAt : 0;
+      const isFresh = Date.now() - updatedAt <= AgentSpawner.ONLINE_THRESHOLD_MS;
+
+      if (!isFresh) return false;
 
       // Case-insensitive check to match router behavior
       const lowerName = name.toLowerCase();
@@ -1094,6 +1105,27 @@ export class AgentSpawner {
     }
   }
 
+  private isAgentRecentlySeen(name: string): boolean {
+    if (!this.registryPath) return false;
+    if (!fs.existsSync(this.registryPath)) return false;
+
+    try {
+      const raw = JSON.parse(fs.readFileSync(this.registryPath, 'utf-8'));
+      const agents = Array.isArray(raw?.agents)
+        ? raw.agents
+        : typeof raw?.agents === 'object' && raw?.agents !== null
+          ? Object.values(raw.agents)
+          : [];
+      const lowerName = name.toLowerCase();
+      const agent = agents.find((entry: { name?: string; lastSeen?: string }) => typeof entry?.name === 'string' && entry.name.toLowerCase() === lowerName);
+      if (!agent?.lastSeen) return false;
+      return Date.now() - new Date(agent.lastSeen).getTime() <= AgentSpawner.ONLINE_THRESHOLD_MS;
+    } catch (err: any) {
+      console.error('[spawner] Failed to read agents.json:', err.message);
+      return false;
+    }
+  }
+
   /**
    * Save workers metadata to disk for CLI access
    */

src/cli/index.ts

@@ -457,7 +457,12 @@ program
         console.log('');
         console.log('Auto-spawning agents from teams.json...');
 
-        spawner = new AgentSpawner(paths.projectRoot, undefined, dashboardPort);
+        spawner = new AgentSpawner({
+          projectRoot: paths.projectRoot,
+          dashboardPort,
+          onMarkSpawning: (name) => daemon.markSpawning(name),
+          onClearSpawning: (name) => daemon.clearSpawning(name),
+        });
 
         for (const agent of teamsConfig.agents) {
           console.log(`  Spawning ${agent.name} (${agent.cli})...`);

src/daemon/orchestrator.ts

@@ -343,7 +343,11 @@ export class Orchestrator extends EventEmitter {
 
     // Ensure spawner exists
     if (!workspace.spawner) {
-      workspace.spawner = new AgentSpawner(workspace.path);
+      workspace.spawner = new AgentSpawner({
+        projectRoot: workspace.path,
+        onMarkSpawning: (name) => workspace.daemon?.markSpawning(name),
+        onClearSpawning: (name) => workspace.daemon?.clearSpawning(name),
+      });
     }
 
     const result = await workspace.spawner.spawn({
@@ -445,7 +449,11 @@ export class Orchestrator extends EventEmitter {
       workspace.status = 'active';
 
       // Create spawner
-      workspace.spawner = new AgentSpawner(workspace.path);
+      workspace.spawner = new AgentSpawner({
+        projectRoot: workspace.path,
+        onMarkSpawning: (name) => workspace.daemon?.markSpawning(name),
+        onClearSpawning: (name) => workspace.daemon?.clearSpawning(name),
+      });
 
       // Set up agent death notifications
       workspace.spawner.setOnAgentDeath((info) => {