crewforth

agent
Security Audit
Fail
Health Pass
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 24 GitHub stars
Code Fail
  • child_process — Shell command execution capability in bin/cli.js
  • spawnSync — Synchronous process spawning in bin/cli.js
  • fs.rmSync — Destructive file system operation in bin/cli.js
  • process.env — Environment variable access in bin/cli.js
  • fs module — File system access in bin/cli.js
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

Enterprise engineering workflow for Claude Code — not just prompts. AI agents that plan, build, audit, and ship with security gates, privacy checks, and approval-controlled commits. Safely adopt it into new or existing repositories.

README.md
Crewforth

Version
License

🇬🇧 English · 🇹🇷 Türkçe

Crewforth is your engineering crew for Claude Code.

It adds subagents, skills, commands and hooks to Claude Code: 12 specialist agents that each own a domain,

39 skills that hold the method, 10 commands you start with /crew-…, and hooks that enforce the rules that matter.

Watch the one-minute overview on crewforth.com

Why Crewforth

  • Work goes to a specialist. An unclear request is planned before any code is written, server work goes to crew-backend-expert, and a risky change is reviewed by crew-security-expert before it closes. A routing hook names the owner beside your request.
  • The rules that matter are enforced by gates, not remembered. A destructive command is refused before it runs, a commit waits for your approval, and a leaked key never reaches history.
  • Every result is measured and published, including the ones that did not hold. The same prompt is run with Crewforth and without it, and graded on what each left on disk. See How we measure.

Quick start

npx crewforth init              # set up a new project
npx crewforth add <agent|skill> # add one agent or skill only
npx crewforth studio            # open the Studio panel

You approve a summary before anything is written, and add copies into ./.claude without the full install. For an existing repository, npx crewforth adopt lands everything on a separate branch, staged and uncommitted; main is never touched. Then run /crew-doctor in Claude Code to confirm the setup.

How a session flows

Command flow: /crew-plan, the specialists, /crew-review, /crew-ship with your approval, /crew-handoff
Step Command What happens
Plan /crew-plan crew-planner turns the request into tasks with acceptance criteria
Build (routed) the owning agent writes the change and applies its skills
Review /crew-review security, privacy, performance and test audits run in parallel
Ship /crew-ship crew-review-agent records a clean review; crew-commit-agent proposes the commit and waits for you
Hand off /crew-handoff handoff writes the state down for the next session

10 commands in all, each started with its /crew-… name: /crew-plan, /crew-review, /crew-ship, /crew-handoff, /crew-brainstorm, /crew-update, /crew-doctor, /crew-gates, /crew-skill, /crew-studio.

The agents

12 specialist agents, each saying who owns the work and when. The method lives in the skills: crewforth.com/skills.

Agent Owns
crew-planner scope and acceptance criteria when a request is unclear
crew-backend-expert server, API and business logic, in any stack
crew-database-expert schema, migrations, indexes and caching
crew-frontend-expert UI, components and client work, on web and mobile
crew-devops-expert deployment, CI pipelines and incidents
crew-security-expert auth, injection and secrets; required for security-critical changes
crew-privacy-agent personal data under KVKK, GDPR and declared regimes
crew-test-expert tests, coverage and regressions
crew-performance-expert hot paths, queries, rendering and payloads
crew-review-agent the code-health review a commit needs
crew-commit-agent the commit proposal, which waits for your approval
crew-session-manager session fill and the handoff

Gates

Rule Enforced by
Commit and push need your approval, in every permission mode guard-bash.sh
A commit needs a clean review of that exact diff guard-bash.sh and the record crew-review-agent writes
Destructive commands (reset --hard, force push, rm -rf, --no-verify) are refused guard-bash.sh
No AI-authorship trace reaches a commit pre-commit and commit-msg git hooks
No API key, token or private key reaches a commit pre-commit secret scan
A gate file cannot be edited or deleted to switch the gate off guard-write.sh

Every hook and rule: crewforth.com/gates. Gates stop accidents, not determined attempts; for a hard boundary, use a devcontainer or a VM.

Studio

Studio is a local panel that draws a delegation as it happens: each agent is a node showing what it is doing, what it has spent and what it reported. It reads every Claude Code session on the machine, opens with /crew-studio or npx crewforth studio, and binds to 127.0.0.1 only. More: crewforth.com/studio.

Install and update

Channel Command
npx npx crewforth init for a new project, npx crewforth adopt for an existing one
Claude Code plugin /plugin marketplace add Crewforth/crewforth, then /plugin install crewforth@crewforth
No Node The GitHub release archive, see crewforth.com/install

Requirements: Claude Code 2.1.214 or later (tested on 2.1.282), and Node.js 20 or later for npx (22 or 24 recommended).

Coming from the 2.x plugin? The plugin and its marketplace were renamed, so a 2.x install does not update to 3.0 on its own. Switch once:

/plugin uninstall claude-starter-kit
/plugin marketplace add Crewforth/crewforth
/plugin install crewforth@crewforth

When a new version is published, Claude asks once, at the start of a session, whether to update now, later or never for that version; it never updates on its own. /crew-update runs the update and reports what changed, and ./CLAUDE.md is never touched. On Windows, use Git Bash. Every option: crewforth.com/install.

How we measure

We run the same prompt in a project with Crewforth and in a bare one, and grade each on what it left on disk. The rule a result must meet is written down before the run. Every result is published with its reasoning, including the ones where that rule did not hold: evals/README.md.

Contributing, licence and links

Issues and pull requests are welcome at github.com/Crewforth/crewforth. A new agent or skill follows the contract in kit/AGENT_TEMPLATE.md, and bash packaging/verify.sh must pass.

MIT, see LICENSE.

Reviews (0)

No results found