AI Skill MartBrowse skillsPromptsProgramsLearnWhat is a Skill?
 
← Back to list

Huashu Agent Swarm (Infinite Agent Loop)

Runs many Claude CLI instances in parallel via tmux and git worktrees, coordinating purely through git for large-scale project development.

AutomationAdvanced★ 2,232⑂ 308AI score 9/10Last updated: Nov 2, 2026

What it does

  • Launches N Claude CLI agents at once with no master orchestrator — coordination happens entirely through the git repository.
  • Each agent claims work via current_tasks/*.lock files, tracks overall progress in TASKS.md, and watches teammates through git log.
  • Uses git worktree so every agent gets an isolated checkout while sharing one .git object store (no duplicated clones).
  • Full lifecycle scripts: initialize (setup_project.sh), launch (start_swarm.sh), observe (web dashboard on port 8420, status.sh), and shut down with branch merge + worktree cleanup (stop_swarm.sh).
  • Lets you steer the running swarm mid-flight via send_input.sh or the dashboard input box (written to HUMAN_INPUT.md).

Who it's for

  • Developers with large, easily parallelized workloads — bulk test writing, sweeping refactors, multi-module implementation.
  • Comfortable-with-tmux/git CLI users who accept the token cost of autonomous agent experiments.
  • Anyone studying practical multi-agent orchestration patterns.

Examples

  1. "Start swarm mode with 8 agents and raise this repo's test coverage to 80%" — each agent locks a different module.
  2. "Parallel-develop this compiler project; the initial backlog is in TASKS.md" — agents claim, implement, and push in a loop.
  3. While running, open http://localhost:8420, review per-agent logs, and broadcast "keep commits small to avoid merge conflicts."

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/alchaincyf/huashu-skills/HEAD/huashu-agent-swarm/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on AI Skill Mart.
Copy the huashu-agent-swarm folder from the GitHub repo alchaincyf/huashu-skills into my ~/.claude/skills/huashu-agent-swarm/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/alchaincyf/huashu-skills.git && mkdir -p ~/.claude/skills && cp -r huashu-skills/huashu-agent-swarm ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Install prerequisites: brew install tmux (macOS), git, Python 3, and a logged-in claude CLI.
  2. Clone the skill repo: git clone https://github.com/alchaincyf/huashu-skills.git
  3. Copy the skill folder into Claude: mkdir -p ~/.claude/skills && cp -r huashu-skills/huashu-agent-swarm ~/.claude/skills/
  4. Make the helper scripts executable: chmod +x ~/.claude/skills/huashu-agent-swarm/scripts/*.sh
  5. Make sure your target project is a clean git repository (run git init and set a remote if needed).
  6. Open Claude Code and ask it to "run swarm mode to develop this project in parallel" to trigger the skill.
  7. For your first run, start with 2–3 agents and a capped session count to gauge API cost and merge behavior, then scale up. Always shut down with stop_swarm.sh.
View source on GitHub ↗License: MIT