What does this skill do, and when should you use it?
This is a contributor-facing development skill for the ClawTeam repository, explicitly not a general end-user usage guide. It ships bootstrap scripts that pin the local clawteam command to a fixed uv environment and symlink the repo's skills into other local projects. It prescribes a layered validation strategy: start with ruff and targeted pytest runs, then escalate to a real CLI smoke test and harness testing only when needed. Use it after changing ClawTeam source to prove a change still works across spawn, task, inbox, and board flows.
- Runs scripts/bootstrap_clawteam_dev.sh to create a fixed uv env at ~/.clawteam-venv, install the repo with dev dependencies, write ~/.local/bin/clawteam, and add a clawteam() shell function to ~/.bashrc
- Runs scripts/link_local_clawteam_skills.sh /path/to/project to symlink ./.agents/skills/clawteam and ./.claude/skills/clawteam in a target project back to this repository
- Runs the smallest validation first: ruff check clawteam/ tests/ plus targeted pytest files, broadening only when a change crosses modules
- Drives a real multi-agent smoke test via clawteam team spawn-team, task create, spawn, board show, and task wait in tmux or subprocess mode
- Validates harness behavior with clawteam harness conduct and cleans leftovers with clawteam team cleanup
- A ClawTeam contributor who changed spawn/task/inbox code and wants minimal validation instead of the full suite
- A developer reproducing a concrete bug at the spawn, board, task, inbox, or harness layer
- A developer who wants another local project to use the repo's current, unpublished ClawTeam skills directly
- A reviewer walking a multi-agent workflow end-to-end against the current checkout
- Smoke-testing the real CLI in tmux or subprocess mode
- End users who want to orchestrate their own projects with ClawTeam — the skill is explicitly repository-development oriented, not a general usage guide; use the user-facing sibling skills instead
- Machines missing the prerequisites (clawteam command, tmux, an installed CLI agent like claude/codex) — smoke tests cannot run without them
How do you install this skill?
- The bootstrap script modifies ~/.bashrc and writes a launcher to ~/.local/bin; back up affected files before running — no official rollback steps are documented.
- The link script links to REPO_ROOT/skills/clawteam; verify this path exists in the current checkout to avoid dangling symlinks.
- SKILL.md version (0.2.0) conflicts with pyproject version (0.3.0); verify the actually installed version before use.
- This is a static review; no scripts were executed, so the assessment does not constitute empirical verification of script behavior.
- Smoke tests depend on tmux and external CLI agents (claude/codex); confirm reachability of related services from mainland-China networks.
- Shell / CLI
- Local filesystem
Python 3.10+ (uv)tmuxclawteam CLIa CLI agent such as claude or codex for real workers
The skill lives at .agents/skills/clawteam-dev/ inside the ClawTeam repository and is used with the repo; there is no standalone install flow. Run the bootstrap on first use: (repo root)
git clone https://github.com/HKUDS/ClawTeam.git
cd ClawTeam
bash scripts/bootstrap_clawteam_dev.sh
source ~/.bashrc
clawteam --versionHow do you use this skill?
Once installed, send your agent any of these to trigger it:
- I changed the task update logic in clawteam — run the minimal validation with this skill.
- Help me reproduce an inbox bug where a spawned worker never reports back to the leader.
- Use harness conduct on the current checkout to validate a one-worker implementation-plan flow.
- Symlink this repo's ClawTeam skills into /path/to/my-project so I can use the local version there.
Use it inside the ClawTeam repo. After code changes, run the smallest validation first: (repo root)
ruff check clawteam/ tests/
pytest tests/<target_file>.py -qWiden the scope only when the change crosses modules:
pytest tests/test_cli_commands.py -q
pytest tests/test_spawn_backends.py -q
pytest tests/test_harness.py tests/test_event_bus.py -qFor a real multi-agent smoke test:
clawteam team spawn-team dev-smoke -d "Local ClawTeam smoke test" -n leader
clawteam task create dev-smoke "Smoke test worker" -o worker1 -p high
clawteam spawn --team dev-smoke --agent-name worker1 --task "Report your status to leader and mark the task completed when done."
clawteam board show dev-smoke
clawteam task wait dev-smoke --timeout 300 --poll-interval 5
clawteam team cleanup dev-smokeFor harness validation:
clawteam harness conduct dev-harness --goal "Create a small implementation plan and execute it with one worker" --cli codex --agents 1What are this skill's strengths and limitations?
- The bootstrap pins clawteam to one fixed uv environment, avoiding command confusion across Python environments
- A clear layered validation policy: minimal by default, broad end-to-end runs only when changes cross spawn/runtime/workflow boundaries
- Complete, copy-pasteable commands spanning unit tests, a real multi-agent smoke test, and harness testing
- Only meaningful inside the ClawTeam repository; useless for any other codebase
- Heavy local prerequisites: Python 3.10+ with uv, tmux, and an installed CLI agent for real workers
- No evidence in the source of dedicated test coverage for this skill itself, and the smoke scripts' behavior across platforms is unverified
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| ClawTeam Repository Development Skill this page | 53 · Use with care | ★ 5.5k | 5mo ago | MIT |
| ClawTeam Multi-Agent Coordination Skill | 54 · Use with care | ★ 5.5k | 5mo ago | MIT |
| ClawTeam Multi-Agent Coordination | 51 · Use with care | ★ 1.5k | 3mo ago | MIT |
| ClawTeam Multi-Agent Swarm Coordination | 49 · Use with care | ★ 1.5k | 3mo ago | MIT |
| ClawTeam Multi-Agent Swarm Skill (Hermes Edition) | 48 · Use with care | ★ 1.5k | 3mo ago | MIT |
Compared with ad-hoc throwaway test scripts, it standardizes environment bootstrap and validation commands; compared with the user-facing sibling skill in this collection, it targets source contributors rather than everyday orchestration. The source names no third-party equivalents.
How did FollowSkills review this skill?
Script behaviors are disclosed in SKILL.md (writes ~/.bashrc, ~/.local/bin, fixed venv); link script errors out on non-symlink targets — decent data-flow disclosure. But bootstrap modifies ~/.bashrc and silently overwrites existing symlinks without explicit user confirmation, and no rollback/uninstall path is documented; deducted.
Bootstrap gives a clear error when uv is missing and handles bashrc block idempotently; however, link script targets REPO_ROOT/skills/clawteam, an unverified path assumption inconsistent with the .agents/skills layout, key-path reproduction cannot be confirmed statically, and the test suite does not cover the skill's own scripts; deducted.
Audience and boundaries are explicit ('repository-development oriented, not a general end-user usage guide'), prerequisites (tmux, uv, CLI agents) listed, trigger semantics clear; but failure/non-fit boundaries are thin and no Chinese-environment notes; deducted.
Versioned (0.2.0), MIT license, well-layered docs; but SKILL.md version 0.2.0 conflicts with pyproject 0.3.0, no changelog or maintenance-ownership statement, hidden path assumptions; deducted.
As a repo-internal dev bootstrap the goal is sound and the bootstrap/link/validate flow is directly usable; but static review cannot verify actual script output, and dependence on external CLIs (claude/codex) makes benefit claims hard to corroborate; deducted.
Repo has CI workflows and a tests/ directory (adapter tests cover the spawn layer), but CI and tests do not cover this skill's scripts; key claims (bootstrap works, link paths correct) rest on author statements without third-party reproduction; deducted.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →