Multi Agent Coordinator

Multi-Agent Coordinator

Coordinate multiple AI agents, discover active sessions, send messages between agents, and delegate work. This skill enables the PopeBot Docker agent to work alongside other agents in a multi-agent workflow.

When to Use

• Task splitting: A large task needs to be divided between multiple agents
• Coordination: Multiple agents need to collaborate on related work
• Status checking: Check what other jobs/agents are currently running
• Context sharing: Pass information from one agent to another
• Delegation: Offload sub-tasks to specialized agents

Architecture

This skill works by interacting with the GitHub API to discover and coordinate with other PopeBot agents:

1. List active jobs: Query GitHub Actions workflows to find running jobs
2. Query job status: Get details about specific jobs
3. Coordinate via git: Use git branches and commits to share context between agents

Usage

List Active Jobs

node /job/.pi/skills/multi-agent-coordinator/list-jobs.js

Shows all currently running GitHub Actions workflow runs.

Check Specific Job Status

node /job/.pi/skills/multi-agent-coordinator/job-status.js <job-id>

Get detailed status of a specific job.

Send Message to Another Agent

node /job/.pi/skills/multi-agent-coordinator/send-message.js <job-branch> "<message>"

Leave a message for another agent by creating a file in their working directory.

Request Help from Another Agent

node /job/.pi/skills/multi-agent-coordinator/request-help.js <target-branch> "<task description>"

Create a help request that another agent can pick up.

Check for Waiting Messages

node /job/.pi/skills/multi-agent-coordinator/check-messages.js

Check if any messages have been left for the current agent.

Coordinate Multi-Agent Task

node /job/.pi/skills/multi-agent-coordinator/coordinate.js --task "<overall task>" --agents 3

Break down a large task and coordinate multiple agents to work on it in parallel.

Output Format

List Jobs Output

{
  "total_count": 5,
  "jobs": [
    {
      "id": "123456789",
      "name": "Run Job",
      "status": "in_progress",
      "branch": "job/abc-123",
      "started_at": "2026-02-25T12:00:00Z",
      "html_url": "https://github.com/owner/repo/actions/runs/123456789"
    }
  ]
}

Job Status Output

{
  "id": "123456789",
  "name": "Run Job",
  "status": "in_progress",
  "conclusion": null,
  "branch": "job/abc-123",
  "run_number": 42,
  "event": "push",
  "created_at": "2026-02-25T12:00:00Z",
  "updated_at": "2026-02-25T12:30:00Z",
  "html_url": "https://github.com/owner/repo/actions/runs/123456789"
}

Common Workflows

Task Delegation

User: Analyze all the log files in /logs and create a summary
Agent: [Splits task] 
  1. Uses list-jobs to check if other agents are available
  2. Uses coordinate to delegate log analysis to multiple agents
  3. Aggregates results from all agents

Multi-Agent Research

Agent: [Coordinate research task]
  1. coordinate.js --task "Research AI agents" --agents 3
  2. Agent 1: Researches OpenClaw
  3. Agent 2: Researches ZeroClaw
  4. Agent 3: Researches AutoGPT
  5. All results combined into final report

Check Existing Work

Agent: [Before starting new work]
  1. list-jobs.js to see what's already running
  2. check-messages.js for any pending requests
  3. Integrates with existing work rather than duplicating

Integration with Other Skills

• With memory-agent: Store coordination history for future reference
• With session-files: Track what each agent has been working on
• With modify-self: Can create new job branches for delegation

Requirements

• GitHub CLI (gh) authenticated with appropriate permissions
• GH_TOKEN environment variable or gh CLI configured
• GH_OWNER and GH_REPO environment variables (optional, auto-detected from git remote)

Limitations

• Only works with PopeBot/GitHub Actions-based agents
• Message passing requires write access to the repository
• Coordination is asynchronous (agents may not be online simultaneously)
• No real-time messaging - uses file-based communication via git

Tips

1. Use descriptive branch names: Makes it easier to identify what other agents are working on
2. Check for existing work: Always list jobs before starting large tasks
3. Leave context: When delegating, include relevant context in messages
4. Monitor completion: Use list-jobs to track when delegated tasks complete