A setup that makes AI-written code safe to merge for a team using Claude Code: shared project rules, permission guardrails that keep Claude away from production and secrets, and a review gate so nothing merges without an independent check.
It's a plugin with twelve agents that work as a software team, from manager and tech lead to QA, security and reviewer. Hooks enforce who may write where and when work counts as done, and commit checks block secrets and client names.
Like this project
Posted Oct 20, 2026
A setup that makes AI-written code safe to merge: shared rules, permission guardrails and a review gate, as a plugin with twelve agents that work as a team.