Agent skill · software engineering · aiskillstore
swarm
Autonomous multi-agent workflow system for complex coding tasks. Use when the user requests non-trivial changes that would benefit from autonomous execution with built-in review gates. Ideal for refactoring modules, adding major features, migrating codebases, adding comprehensive test coverage, or any multi-step development task that can run autonomously. NOT for simple fixes, typos, or single-file changes.
Why this skill is useful
Adds an executable swarm script that enables autonomous multi-agent workflows for complex coding tasks, which the AI cannot generate on its own.
What it needs
Requires bash installed locally. About 5k tokens when loaded. Last updated 2026-08-07. 405 stars on the source repository.
What this skill does
Swarm: Autonomous Multi-Agent Workflows Execute complex coding tasks autonomously using multiple agents working from a coordinated plan. How It Works 1. Interview — Gather requirements and context from user 2. Plan — Create structured markdown plan with checkboxes and review gates 3. Execute — Spawn autonomous agents to work on the plan 4. Complete — Agents signal completion with .done files Workflow Step 1: Interview the User When this skill is invoked, gather the information needed to create an effective plan: Required information: Goal: What does "done" look like? (Be specific and measurable) Context: What files/directories are involved? Key constraints? Test strategy: What commands verify success? (e.g., npm test, pytest) Plan location: Where should the plan file be created? Interview questions: Ask follow-up questions as needed to clarify scope, constraints, or success criteria. Step 2: Create the Plan Use the template at assets/PLANTEMPLATE.md as a starting point. Customize it based on the user's requirements. Plan structure: ## Goal — One sentence defining completion criteria ## Context — Codebase structure, key files, constraints ## Plan — Phases with checkboxes for each step ## Agent Roles — Review gates (Reviewer, Simplifier, Tester, Gatekeeper) ## Rules — Behavioral constraints for agents ## Log — Communication space for agents across iterations Critical components: 1. Checkboxes — Every step must be checkable. Avoid prose. 2. Review Gates — Force perspective shifts at intervals: Reviewer (every 5th iteration): Fresh eyes on progress vs goal Simplifier (before final phase): Identify overengineering Tester (before final phase): Write and run tests Gatekeeper (before .done): Mandatory checklist before completion 3. Gatekeeper Checklist — Must include: Commands to run with output (npm test, git diff --stat) Explicit yes/no questions with evidence GO/NO-GO decision with reasoning 4. Log Section — Agents append work here. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills aiskillstore/swarm