Agent Delegation (Inter-Agent ONLY)
Delegate tasks to other AI agents in the Content Factory ecosystem via N8N WF 26 Agent Dispatcher.
CRITICAL: When to Use vs When NOT to Use
USE this skill when:
- User asks you to delegate reasoning to another agent
- "Ask Moltis to analyze competitive landscape"
- "Have TG-Kombain run a full parsing strategy for channel X"
- Task requires another agent's judgment, not just data retrieval
DO NOT use this skill when:
- User wants to publish content → use
n8n-apiskill - User wants data or stats → use
tg-kombainskill - User wants to trigger a workflow → use
n8n-apiskill - User wants system health check → use
tg-kombainskill
Rule of thumb: If the request can be fulfilled by a single API call, don't delegate. Only delegate when the task needs multi-step reasoning from another agent.
Architecture
You (IDEA/Secretary) ──► N8N WF 26 ──► Target Agent
├── Moltis Atlas (deep research, analysis)
├── TG-Kombain (complex multi-step automation)
└── N8N (workflow orchestration)
When to Delegate to Each Agent
| Agent | Delegate for... | NOT for... |
|---|---|---|
| Moltis | Deep research, competitive analysis, complex reasoning, monitoring strategy | Simple data queries (use tg-kombain) |
| TG-Kombain | Complex multi-step automation requiring planning | Single API calls (use tg-kombain) |
| N8N | Complex workflow orchestration requiring judgment | Simple workflow triggers (use n8n-api) |
Delegation via N8N Webhook
Endpoint: POST $N8N_BASE_URL/webhook/agent-dispatch
Body (AgentMessage):
{
"from_agent": "openclaw",
"to_agent": "moltis",
"action": "delegate",
"task_id": "unique-task-id",
"payload": {
"message": "Describe the task clearly..."
},
"metadata": {
"priority": "normal",
"timeout_seconds": 300,
"callback_url": "https://openclaw-v3-production.up.railway.app/hooks/agent"
}
}
Valid Agents
openclaw(this agent — you)moltis(deep research, Rust sandbox, 79 MCP tools: 76 TG-Kombain + 3 N8N)tg-kombain(Telegram automation platform)n8n(workflow orchestration)
Valid Actions
delegate— ask another agent to perform a reasoning taskquery— ask for synthesized analysis (not raw data)notify— send a notification (no response expected)result— return task result (callback from target agent)
Example: Delegate Research to Moltis
curl -s -X POST "$N8N_BASE_URL/webhook/agent-dispatch" \
-H "Content-Type: application/json" \
-d '{
"from_agent": "openclaw",
"to_agent": "moltis",
"action": "delegate",
"payload": {
"message": "Analyze competitive landscape for Telegram channels in crypto niche. Compare top 10 channels by engagement, growth, and content strategy. Use get_audience_insights and get_parsed_channels tools."
},
"metadata": {
"priority": "normal",
"timeout_seconds": 600,
"callback_url": "https://openclaw-v3-production.up.railway.app/hooks/agent"
}
}'
Result Callback
When a delegated task completes, the result arrives at your hooks endpoint:
POST /hooks/agent with body:
{
"task_id": "...",
"from_agent": "moltis",
"status": "completed",
"result": {
"summary": "Analysis complete...",
"data": {...}
}
}
The result is automatically delivered to the active conversation session.
Anti-Loop Protection
Each message carries a hop_count in metadata. The system rejects messages with hop_count >= 3 to prevent infinite delegation loops.
Converted and distributed by TomeVault — claim your Tome and manage your conversions.