Stats
Actions
Available In
Tags
Plugin
crew
The new way to build: you bring the vision, the crew builds it — auditing and simplifying every commit as it lands, closing each chunk with a cross-model Codex pass and each turn with a /code-review gate. Fast iteration in, hardened code out.
What's Inside
Skills4
auditor
/auditor
Correctness-only review of a committed diff range — also runnable on its own. Pass a git range (e.g. HEAD~1..HEAD or origin/main..HEAD) plus optional intent via $ARGUMENTS.
editor
/editor
Editorial pass over a committed diff's written content — code comments and authored prose. Cuts narration, conversational residue, and filler while keeping what the reader's task needs. Also runnable on its own. Pass a git range (e.g. HEAD~1..HEAD) plus optional intent via $ARGUMENTS.
flow
/flow
The default way to do any work that writes or changes code or durable artifacts — features, refactors, fixes, design docs, investigations: build it through this loop rather than editing freely. Skip it only for pure conversation, read-only exploration, or a one-line typo — and if you're unsure whether work qualifies, it does; invoke it. Pass an optional unit/chunk label via $ARGUMENTS.
specialist
/specialist
The crew's line to OpenAI Codex — a different model you bring in for an outside read: review a committed diff for correctness/security/contract bugs, or get fresh eyes on a stubborn bug or design call. Pass a git range or a question via $ARGUMENTS.
README
The new way to build.
Fast iteration in, hardened code out.
crew is a Claude Code plugin. You don't write the code — you say what to build, and the crew builds it: reviewing every commit as it lands and bringing in a second model to harden it, so what ships actually holds.
Meet the crew
/crew:flow | the loop | drives build → review for every commit, chunk, and turn |
/crew:auditor | the correctness pass | logic errors, edge cases, broken contracts, intent drift |
/crew:editor | the written-content pass | strips AI-isms, narration, and conversational residue |
/crew:specialist | the outside read | brings in OpenAI Codex — a different model — to re-check the work |
flow runs the rest of the crew for you — or call them in yourself: auditor
and editor on any diff, specialist for a cross-model review or fresh eyes on
a stubborn bug.
How it works
Every commit the crew makes is audited for correctness and run through
/simplify for quality the moment it lands. Each chunk then closes with a
cross-model pass — the specialist brings in OpenAI Codex, a different
model, to catch what Claude missed — and the turn ends with a full
/code-review on top.
That second model is the point: crew never lets one model grade its own homework. What ships has been audited, simplified, edited, and cross-examined.
Install
/plugin marketplace add Jud/crew
/plugin install crew@crew
Then, on any commit-producing work:
/crew:flow
One prerequisite. The specialist runs the real OpenAI Codex CLI. The bridge
to it installs with crew, but the CLI is yours to provide — the codex command
on your PATH, plus OpenAI auth. Without it, the cross-model pass skips and
everything else still runs.
License
MIT
vibe with your crew
Stats
Version1.0.0
LanguageHTML
Stars0
MaintenanceExcellent
LicenseMIT
Last Commit
Added
Dependencies
skill-codex
Tags
