huashu-agent-swarm

You can coordinate multiple autonomous agents by using a Git-based, self-organizing swarm with per-agent worktrees. This setup isolates changes, uses task locks for clear ownership, and tracks progress through commit history for auditable parallel coding.

Git-based task management works by having agents independently select and execute tasks from TASKS.md. Agents claim work via current_tasks lock files, operate in isolated Git worktrees, and commit changes to ensure distributed coordination without central orchestration.

Yes, you can use Git worktrees to isolate changes for parallel coding tasks. The swarm uses per-agent worktrees so each autonomous agent can independently develop, execute, and commit tasks without interfering with others.

The best way to monitor autonomous agent progress is through a lightweight dashboard and simple logs. These tools provide visibility into agent status, progress tracking, and debugging logs for peer collaboration within the swarm.

No, you do not need a central master to run a multi-agent swarm. The system uses a self-organizing swarm where each agent independently selects and executes tasks, suitable for large projects requiring minimal central orchestration.

You should avoid using a Git-based swarm if your large project requires strict central orchestration rather than distributed task ownership. The self-organizing swarm relies on minimal central control and independent agent execution via lock files and worktrees.