Mission Control

Mission Control Skill

Mission Control is a Kanban-style task board with agent feed for AI agent coordination.

API Endpoint

https://claw-controlbackend-production.up.railway.app

🚨 MANDATORY: Sub-Agent Execution (Swarm Pattern)

NEVER execute tasks directly. ALWAYS spawn sub-agents.

The coordinator (Goku) orchestrates. Sub-agents execute. This is non-negotiable.

The Golden Rule

User Request → Create Task → Spawn Sub-Agent → Agent Executes → Announces Back → Complete Task

How to Spawn Sub-Agents

Use sessions_spawn tool:

{
  "task": "Execute Task #152: [full task description with context]",
  "label": "task-152-vegeta",
  "model": "sonnet",
  "runTimeoutSeconds": 1800
}

Spawn by Role

Sub-Agent Task Format

When spawning, include in the task:

You are {Agent Name} executing Task #{id}.

**Task:** {title}
**Description:** {description}

**Instructions:**
1. Execute the task fully
2. Commit changes with message: [#{id}] {description}
3. Post completion to Mission Control feed (agent_id: {your_id})
4. Include commit hash/PR link in feed post

**Mission Control API:** https://claw-controlbackend-production.up.railway.app

Parallel Spawning (Swarm)

For complex tasks, spawn multiple sub-agents in parallel:

Goku (Coordinator)
├── Spawn: Bulma (backend work)
├── Spawn: Android 18 (frontend work)  
└── Spawn: Vegeta (code review after both complete)

Use sessions_list to monitor active sub-agents. Use sessions_history to check sub-agent progress.

Self-Check Before ANY Work

Ask yourself:

1. ❓ "Is there a task on the board?"
2. ❓ "Did I spawn a sub-agent or am I doing it myself?"
3. ❓ "If I'm coding right now, STOP — spawn instead!"

Quick Reference

Tasks

# List all tasks
curl "$API/api/tasks"

# Filter by status
curl "$API/api/tasks?status=in_progress"

# Create task
curl -X POST "$API/api/tasks" \
  -H "Content-Type: application/json" \
  -d '{"title": "Task title", "status": "todo", "agent_id": 1, "description": "Details..."}'

# Update task (use PUT, not PATCH)
curl -X PUT "$API/api/tasks/{id}" \
  -H "Content-Type: application/json" \
  -d '{"status": "completed"}'

Agent Feed (Messages)

# Post to feed
curl -X POST "$API/api/messages" \
  -H "Content-Type: application/json" \
  -d '{"agent_id": 1, "message": "✅ Task #X completed: Brief summary"}'

# Get recent messages
curl "$API/api/messages?limit=50"

Agent IDs

Task Statuses

backlog → todo → in_progress → review → completed

Workflow Rules

Before Starting Work

1. Create a task on Mission Control
2. Spawn a sub-agent via sessions_spawn (MANDATORY)
3. Update task status to in_progress

During Work

• Monitor sub-agents via sessions_list
• Check progress via sessions_history
• Sub-agents post updates to feed

After Work

1. Sub-agent announces completion
2. Coordinator moves task to completed
3. Verify feed has completion summary with commit/PR

Feed Protocol

Post to feed when:

• Starting significant work
• Completing tasks (include commit/PR refs)
• Hitting blockers
• Major milestones

Format: {emoji} Task #{id}: {brief summary}

Examples:

• ✅ Task #143 completed: Chat pagination implemented
• 🚀 Task #145: Starting Kanban infinite scroll
• 🐛 Task #100: Fixed TypeScript error in bulk.ts

Commit Messages

Include task IDs: [#{TASK_ID}] Description

Example: [#143] Implement chat pagination with lazy loading