swarm_task
Decompose a high-level goal into parallel sub-tasks executed by autonomous agents, then synthesize outputs into a single deliverable.
Instructions
Auto-decompose a high-level goal into N parallel sub-tasks, fan out agents across all of them, then run a synthesis agent that produces one unified deliverable. Returns immediately with swarm_id and sub_job_ids. Use get_swarm_status to poll progress.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| goal | Yes | High-level goal to decompose and execute across parallel agents | |
| repo_url | Yes | GitHub repo URL for all sub-agents and the synthesis agent | |
| max_agents | No | Maximum number of parallel sub-agents (default 10, hard cap 50) | |
| agent_model | No | Model override for all spawned agents (optional) | |
| agent_driver | No | Driver for all spawned agents (optional, default: claude) | |
| synthesis_output | No | Path in the repo where the synthesis agent writes its deliverable (default: swarm-synthesis.md) | |
| synthesis_prompt | No | Custom instruction for the synthesis agent. Defaults to: review all outputs and write a unified deliverable. | |
| max_budget_per_agent | No | Max USD budget per agent (sub-agents and synthesis). Default 5. |