fltman

Agent Team — AI skill for Claude Code

AI community

A Claude Code skill that runs a multi-model agent team on a GitHub repo.

How to install Agent Team

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

What Agent Team does

A Claude Code skill that runs a multi-model agent team on a GitHub repo:

Alternatives in AI

Open Multi Agent — TypeScript multi-agent orchestration engine — one runTeam() call from goal to result 5.8k ★
Multi Execute — 执行 - 多模型协作执行 (Multi-Model Collaborative Execution) 1.9k ★
Multi Model Consult — Multi-Model Consultation Skill 1.9k ★

README

agent-team

A Claude Code skill that runs a multi-model agent team on a GitHub repo: Product Owner and Architect on Fable 5.1, Coder on Opus 5, Reviewer on Codex (GPT). The team lives **in the project** as a pinned git submodule. One clone per role, one terminal per clone, all communication through git. Status is derived, never written. The human merges.

See [SKILL.md](SKILL.md) for the design and [team/TEAM.md](team/TEAM.md) for the rulebook the agents get.

New project

One command, from nothing to a running workspace:

curl -sL https://raw.githubusercontent.com/fltman/agent-team/main/new-project.sh | bash -s -- myproj

It creates the GitHub repo if needed (private by default, `--public` to change), makes `./myproj-team/`, clones `po/`, pins the newest tagged team version as a submodule, scaffolds, pushes, clones `architect/`, `coder/`, `reviewer/`, writes `./team` — and opens the four roles, each in its own window: a 2×2 grid of Terminal windows on macOS, a 2×2 tmux session elsewhere (`--no-open` to skip). From a clone of this repo, `./new-project.sh myproj` does the same.

cd myproj-team
./team open        # (again) PO top-left, Architect top-right, Coder and Reviewer below; --tmux for tmux anywhere
                   # every role starts working on launch; Architect/Coder/Reviewer wait on the board and wake up when it's their turn
./team board       # the board, refreshed every 30 s — park it in a corner
./team po          # or start any single role by hand — PO is the one you talk to; ask it to fill in Project conventions first
./team status      # the board, once
./team log         # who did what
./team add coder   # more throughput: coder-2/, started with ./team coder 2

`./team` is a three-line shim; the launcher itself lives in the submodule, so it upgrades with the team.

By hand, the same thing is: `gh repo create`, `git clone … po`, `git submodule add … .claude/skills/agent-team`, then `/agent-team init` inside `po/`.