DimaMuhannad

Multi Agent Coordination Skill — Git skill for Claude Code

Git community

Claude Code Skill that sets up a battle-tested multi-session coordination scaffold — role zones, append-only decision journal, cold-start budget hook, and git-based KPI — for projects run by several p.

How to install Multi Agent Coordination Skill

This entry records only its repository, not the path inside it, so there is no exact command to give. Open DimaMuhannad/multi-agent-coordination-skill and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Multi Agent Coordination Skill does

Alternatives in Git

Epic Sync — Sync epic issue bodies, labels, and local coordination snapshots from GitHub 243.5k ★
Husky — by evmts - Sets up and manages Husky Git hooks by configuring pre-commit hooks, establishing commit message st 440 ★
MCP Agent Mail Rust — Rust MCP server for multi-agent coordination: 34 tools, Git-backed archive, SQLite indexing, advisory file loc 65 ★

README

multi-agent-coordination-skill

A [Claude Code Skill](https://docs.claude.com/en/docs/claude-code/skills) that sets up a coordination scaffold for projects where several agent sessions — each holding a distinct, continuing role/zone — share one repository over an extended period.

The scaffold is platform-agnostic: a role is held by whatever session reads its file and makes commits, so Claude Code, Gemini via Antigravity, Cursor or Aider can each hold one. See [`docs/ru/CROSS_PLATFORM_BRIDGE.md`](docs/ru/CROSS_PLATFORM_BRIDGE.md).

That is enforced by the file layout rather than left to good intentions: the project's rules live in **`AGENTS.md`**, the cross-vendor convention governed since December 2025 by the Linux Foundation's [Agentic AI Foundation](https://www.linuxfoundation.org/press/linux-foundation-announces-the-formation-of-the-agentic-ai-foundation), and `CLAUDE.md` is a four-line adapter that imports it and adds the Claude Code specifics (`claude -n `, `claude agents --json`, `.claude/rules/` globs). Claude Code reads `CLAUDE.md` and not `AGENTS.md`, which is exactly why the adapter exists.

It does not compete with a tool's own multi-agent features. Claude Code's agent teams keep their shared task list in `~/.claude/`, scoped to one session and deleted when it ends; this scaffold is the durable half — in git, on every machine, readable by any tool.

It's not a framework or a running service: there's no server, no live process registry, no database. It's a set of markdown templates and two small stdlib-only Python scripts, distilled from a real multi-role project that ran this paradigm for weeks and fixed the specific problems it hit along the way. Those problems — and why the fix looks the way it does — are written up in [`references/rationale.md`](references/rationale.md).

What's in it

coordination/ templates — role zone files, an append-only decision journal (QUESTIONS.md), cross-role request log (HANDOFFS.md), project-wide a