Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add okqixiaobao727-design/agentcrew-dev-skills --skill crewgit clone --depth 1 https://github.com/okqixiaobao727-design/agentcrew-dev-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/okqixiaobao727-design/agentcrew-dev-skills/crew)<a href="https://agentmods.dev/skills/okqixiaobao727-design/agentcrew-dev-skills/crew"><img src="https://agentmods.dev/badge/skills/okqixiaobao727-design/agentcrew-dev-skills/crew/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/okqixiaobao727-design/agentcrew-dev-skills/crew"><img src="https://agentmods.dev/badge/skills/okqixiaobao727-design/agentcrew-dev-skills/crew.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00037 | $0.01747 |
| Opus 5.5 | $0.00015 | $0.00699 |
| Sonnet 5.5 | $0.00007 | $0.00349 |
| Haiku 4.5 | $0.00004 | $0.00175 |
Grade A, and why
crew scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 118 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Crew
/crew <run-dir> runs the tickets of the run directory /route staged, unattended, and hands back
an integration branch to review. Typing it is the run's whole sign-off: the wave
table the driver validates from each ticket's ## Routing section is the sole routing authority
(ADR-0003).
Your work is judgment: rule on what a child asks, and on what the driver's rule table has no row for. The driver runs the rest — preflight, dispatch, receipts, settlement, merges, tracker closes, advancement, the report — in a tmux window of its own, costing you no turn while it works (ADR-0010).
Contract
Terms are the glossary's (docs/glossary.md).
Reversibility is the authority boundary. You may approve any reversible action — production, remote, deployment, external-service and acceptance runs included — and must record the action's exact undo inside the ruling that approves it. Whether a ticket's acceptance runs at all is your ruling too; its credentials come from the project's own convention, and one the run lacks is parked. An action with no credible undo is parked for the human. The red-line hook blocks known destructive command shapes; this rule stands where the hook misses.
Judgment material is read whole; facts are delegated — read <run-dir>/spec.md whole at run
start, before the first ruling: staging generated it as this run's Reference index to the
maintainer-authored Markdown the coordinator may read whole. The hook compares physical locations.
Code, tests and git remain facts for the witness brief. Every escalation carries the Witness line
that checks its pointers; you run it yourself, before you rule, as
references/triage.md says. When a ruling needs a fact the table lacks,
ask the same Witness directly:
python3 <crew-skill-dir>/assets/witness.py ask \
--run <run-dir> --ticket <NN> --question '<one factual question>'
What ships with it
43 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- assets/accounts.py 11 KB runs code
- assets/advance.py 15 KB runs code
- assets/bounded_read.py 25 KB runs code
- assets/codex/codex_bridge.py 38 KB runs code
- assets/codex/stub_codex.py 11 KB runs code
- assets/codex/test-codex-bridge.sh 59 KB runs code
- assets/coordinator_control.py 13 KB runs code
- assets/dispatch/dispatch.py 52 KB runs code
- assets/dispatch/templates/shapes.toml 22 KB
- assets/dispatch/tests/stub_claude.py 3.0 KB runs code
- assets/dispatch/tests/stub_codex_bridge.py 1.8 KB runs code
- assets/dispatch/tests/stub_tmux.py 2.5 KB runs code
- assets/dispatch/tests/test_dispatch.py 128 KB runs code
- assets/driver/driver.py 217 KB runs code
- assets/driver/tests/harness.py 43 KB runs code
- assets/driver/tests/stub_claude.py 5.5 KB runs code
- assets/driver/tests/stub_codex_bridge.py 4.9 KB runs code
- assets/driver/tests/stub_gh.py 5.2 KB runs code
- assets/driver/tests/stub_review_bridge.py 692 B runs code
- assets/driver/tests/stub_tmux.py 11 KB runs code
- assets/driver/tests/test_clear.py 21 KB runs code
- assets/driver/tests/test_driver.py 297 KB runs code
- assets/driver/tests/test_sweep.py 17 KB runs code
- assets/launch/launch.py 23 KB runs code
- assets/launch/tests/stub_driver.py 6.3 KB runs code
- assets/launch/tests/stub_tmux.py 2.4 KB runs code
- assets/launch/tests/test_launch.py 50 KB runs code
- assets/machine_log.py 91 KB runs code
- assets/merge_driver.py 28 KB runs code
- assets/monitor-wave.sh 6.4 KB runs code
- assets/monitor/monitor.py 126 KB runs code
- assets/monitor/tests/stub_claude.py 2.8 KB runs code
- assets/monitor/tests/stub_tmux.py 2.6 KB runs code
- assets/monitor/tests/test_monitor.py 206 KB runs code
- assets/red-line.sh 8.2 KB runs code
- assets/run_plan.py 54 KB runs code
- assets/settings.local.json 583 B
- assets/test-monitor-wave.sh 11 KB runs code
- assets/test-path-handling.sh 5.2 KB runs code
- assets/tracker.py 14 KB runs code
- assets/witness.py 49 KB runs code
- assets/worktree-guard.sh 1.9 KB runs code
- references/triage.md 13 KB
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed · +4 lines fe52f8b19af9
- 1mo ago Changed · +2 lines 43683acbf3ca
- 1mo ago Changed · +12 lines cd9672d77f0d
- 1mo ago First seen · 100 lines · 37 tokens per session scan A 94bf30af670d
crew is a skill published in the GitHub repository okqixiaobao727-design/agentcrew-dev-skills (1 stars, last pushed 1mo ago), licensed MIT. It adds 37 tokens to every session and 1,747 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
review-switch
Dispatch a review to the Bridge, which resolves the reviewer Lane from this machine's Machine Config. Use when a review is asked for without naming a lane, or when a review-skill invocation was refused and pointed here.
feature-execution
A feature-delivery coordinator that divides a feature into waves of work, assigns agents, manages review rounds, and commits each wave.
done
A work-session finalization process that compares documentation with recent Git changes, updates project knowledge, and can optionally archive a completed feature. Git is a system for tracking changes to code and related files.
init-project
Initialize project with template, git, and GitHub.
agent-framework-py-release
Use when cutting a Python release for the microsoft/agent-framework monorepo. Triggers on "bump py versions", "cut a python release", "prepare release PR for python", "release py packages", "bump python to X.Y.Z", or similar requests to bump Python package versions and prepare a release PR. Handles all four lifecycle…
verify-behavior
Verify or reproduce visible product behavior by driving the real UI with pi-computer-use's checked tools, requiring verified expect postconditions and durable state evidence for meaningful UI flows. Use when triage needs visual reproduction, implementation needs behavioral proof, review needs interactive confirmation…