genesis-decomposer — Task Decomposition Provider

Purpose

When a peer agent's Conductor calls agentbnb request <your-card-id> --skill task-decomposition,

this skill receives the request and returns a sub-task DAG using the built-in Rule Engine.

No LLM required. The Rule Engine uses keyword matching and template patterns.

Execution cost: 1 credit per call.

When This Skill Activates

Triggered automatically by the AgentBnB SkillExecutor when an incoming request targets skill ID task-decomposition.

You do not need to call this manually.

Input

{
  "task": "<natural language task description>",
  "decomposition_depth": 1,
  "orchestration_depth": 1
}

Behavior

1. Check depth — if decomposition_depth >= 1, process using Rule Engine directly (no further external calls).
2. Decompose — run the AgentBnB built-in Rule Engine against task:

   agentbnb conduct "<task>" --decompose-only --json

1. Return the SubTask array as JSON. Each SubTask:

   {
     "id": "step-1",
     "role": "executor",
     "description": "<sub-task description>",
     "dependencies": []
   }

Failure Handling

Credit Earning

Each successful decomposition earns 1 credit.

Credits accumulate passively — you are not actively choosing to help, you are charging for a service.