ClawTeam Multi-Agent Coordination
Turn multiple AI coding agents into a coordinated team via the clawteam CLI: spawn workers, delegate tasks, exchange messages, and monitor boards with git worktree isolation.
Docs disclose data dir, snapshot/restore, atomic writes, and a --no-skip-permissions override, so risks are visible; but spawn defaults to skip-permissions on, inbox receive is destructive by default, and worktrees/context injection happen automatically — least privilege and user confirmation are incomplete, hence deduction.
SKILL.md, CLI reference and workflows are mutually consistent with aligned commands, options and data models; the repo has CI and tests covering spawn adapters. However, static review cannot reproduce key paths, and error-handling/failure-feedback evidence is thin, so the static cap of 10 applies.
Trigger conditions are concrete (spawn agents, task board, multi-agent coordination); inputs, prerequisites (tmux, git, a CLI agent) and non-fit boundaries are clear; China-reachable presets (moonshot-cn, minimax-cn) and Chinese docs exist. Near full marks, but no executed verification.
Layered docs (SKILL.md → references), MIT license, version field, known limitations and notes are present; maintainer HKUDS is identifiable. But SKILL.md says 0.3.1 while pyproject says 0.3.0 — version inconsistency — and no changelog/update path ships with the skill, hence deduction.
README shows multi-scenario use cases and a clear orchestration value claim; however no statically verifiable actual outputs exist (results appear only as narrative logs and video links), so below the static cap of 7.
The repo contains GitHub Actions CI (multi-OS, multi-Python) and committed tests covering spawn adapters and skip-permissions mapping — auditable primary material; but tests do not cover the full key paths the skill advertises (task board, inbox, profiles), so the static cap of 5 applies.
- spawn enables skip-permissions by default, meaning spawned agents bypass permission confirmations; always use --no-skip-permissions or set skip_permissions=false and run only in trusted environments.
- inbox receive is destructive; prefer inbox peek in scripts to avoid losing messages.
- Skill doc version (0.3.1) mismatches package metadata (0.3.0); verify the actual released version before installing.
- This is a static review with no execution; reliability/effectiveness conclusions are inferred from docs and test code — validate on a small scale before production use.
- Depends on external CLI agents (claude/codex etc.) and tmux; the skill will not work if any prerequisite is missing.
What does this skill do, and when should you use it?
ClawTeam is a framework-agnostic CLI tool that organizes multiple AI agents (Claude Code, Codex, Gemini, Kimi, nanobot, OpenClaw) into a collaborative team. A leader agent spawns workers via clawteam spawn, and each worker automatically gets its own git worktree, tmux window, and identity. Teams share a task board with dependency chains and priorities, a file-based inbox messaging system, and monitoring via tiled tmux views, live dashboards, and a Web UI. All state lives as JSON files in ~/.clawteam/ with no database or server required. This skill (SKILL.md v0.3.1) gives agents comprehensive guidance for driving the CLI.
Creates teams with clawteam team spawn-team and spawns worker agents with clawteam spawn (default: tmux backend, claude command, git worktree isolation, skip-permissions on). Manages a shared kanban with dependencies and priorities via task create/update/list/wait. Exchanges messages between agents with inbox send/broadcast/receive/peek. Configures non-default model providers (Moonshot, MiniMax, etc.) through preset and profile commands. Monitors via board show/live/attach/serve; checks cross-worktree conflicts before merges with context conflicts/inject; snapshots and restores team state with team snapshot/restore. All commands support -- output; data is stored under ~/.clawteam/.
- A developer using Claude Code needs to implement multiple features in parallel (API, frontend, tests) and wants a leader agent to split the work and spawn worker agents automatically
- An engineer refactoring a large codebase needs multiple agents working in isolated git worktrees that can later be merged cleanly
- A team using a non-default provider (e.g., Claude Code via Moonshot Kimi) wants reusable runtime profiles generated from presets instead of manual env vars
- Automation workflows where workers should keep polling tasks and inbox after finishing their first task instead of exiting immediately
- A team lead who wants to verify whether parallel agents' changes conflict before merging, using context conflicts
What are this skill's strengths and limitations?
- Framework-agnostic: any CLI agent that can run shell commands can participate, with an auto-injected coordination prompt and no custom SDK
- Git worktree isolation keeps parallel agents conflict-free with inspectable, mergeable branches
- Pure filesystem storage (~/.clawteam/ JSON files) — no Redis, database, or cloud dependency
- Task dependency chains auto-unblock on completion; priorities and -- output support scripting
- Operational tooling built in: snapshots/restore, conflict checks, plan approval, graceful shutdown
- Depends on tmux, a git repo, and a CLI agent that already works standalone — spawn cannot fix a broken agent CLI
- Spawn defaults to skip-permissions; you must explicitly pass --no-skip-permissions to disable, which has security implications
- No test suite or cross-platform validation evidence in the skill docs; Cursor support is labeled experimental at the repo level
- inbox receive is destructive; misuse can lose messages
- Docs give no cost figures — running many parallel agents multiplies model API consumption
How do you install this skill?
- Ensure Python 3.10+, tmux, and at least one CLI coding agent (e.g., claude) are installed. 2. Check with
clawteam --version; if missing, runpip install clawteam(for P2P transport:pip install "clawteam[p2p]"). 3. The skill file lives at docs/skills/clawteam/ in the repo; the README notes install scripts place the skill into detected client skill directories (e.g., ~/.claude/skills/clawteam) — consult the repo for exact installation steps.
How do you use this skill?
Set identity env vars, create a team, and delegate:
export CLAWTEAM_AGENT_ID="leader-001"
export CLAWTEAM_AGENT_NAME="leader"
export CLAWTEAM_AGENT_TYPE="leader"
clawteam team spawn-team my-team -d "Project team" -n leader
clawteam task create my-team "Design system" -o leader
clawteam spawn --team my-team --agent-name worker1 --task "Implement the auth module"
clawteam board attach my-team
Or trigger it in an agent conversation: prompts containing "create a team", "spawn agents", "coordinate multiple agents", or tasks complex enough to split (e.g., "build a full-stack app", "implement multiple features in parallel") activate the skill. Note that inbox receive consumes messages — use inbox peek for non-destructive reads.
How does this skill compare with similar options?
The README contrasts ClawTeam with other multi-agent frameworks: those target humans writing orchestration code and require Docker, cloud APIs, or Redis, and are framework-specific; ClawTeam lets AI agents self-organize via CLI commands, needs only a filesystem and tmux, supports any CLI agent, and uses git worktrees instead of containers for isolation.