swarm

Coordinate N agents on a shared SQLite-backed task queue.

28|5|Updated Jan 26, 2026
One-click install
npx skills add https://github.com/MeroZemory/oh-my-droid --skill swarm-merozemory
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: swarm
Source: https://github.com/MeroZemory/oh-my-droid/tree/main/skills/swarm
Command: npx skills add https://github.com/MeroZemory/oh-my-droid --skill swarm-merozemory

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Coordinates multiple Factory Droid agents on shared work without duplicate effort by managing a single SQLite-backed claim list with heartbeat oversight and deterministic recovery.

Core Features & Use Cases

  • Atomic task claiming via SQLite transactions ensures only one agent owns a task at a time, even under high contention.
  • Heartbeat monitoring and lease timeouts automatically release stalled work, keeping progress moving and preventing deadlocks.
  • Worker preamble and background agents maintain consistent tooling, prohibit recursive spawns, and report absolute paths for reliability.
  • Use Cases include launching executors to fix TypeScript errors, designers to implement responsive layouts, security reviewers to audit APIs, or writers to add documentation in parallel.

Quick Start

Run /swarm 5:executor "fix all TypeScript errors" to launch the coordinated worker pool.

Frequently Asked Questions about swarm

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I coordinate parallel agents to prevent duplicate work in a shared codebase?

Parallel agent coordination is managed via a SQLite-backed task queue that uses atomic claiming logic to ensure only one agent owns a specific task at a time. This prevents multiple agents from performing duplicate edits or conflicting work simultaneously.

What happens if an agent stalls during task orchestration?

Task orchestration includes heartbeat monitoring and lease timeouts that automatically release stalled work. If an agent stops sending heartbeats, its lease expires, returning the task to the queue for another worker to claim deterministically.

How do I launch multiple Factory Droid agents for concurrent bug fixes and documentation work?

You launch concurrent Factory Droid agents by running the swarm command with a specified worker count and task description. For example, /swarm 5:executor launches a coordinated pool of five workers to handle assigned tasks in parallel.

Do I need SQLite to run multi-agent workflows with heartbeat monitoring?

Yes, SQLite session storage is required to maintain task ownership, atomic claiming, and deterministic recovery. The heartbeat monitoring and lease timeout mechanisms depend on this shared database backend to function reliably.

What are the limitations of using a SQLite-backed task queue for parallel agent coordination?

The system relies on SQLite transactions for atomic claiming, which may face contention under extremely high concurrency. A worker preamble is required to prohibit recursive spawns, and agents must report absolute paths for reliable operation.