Runs automated code reviews before commits with a mandatory ADR gate that blocks non-trivial changes unless they're documented in architecture decision records. You can pick between Claude, OpenAI Codex, or Google Gemini as the review engine, or run multiple engines simultaneously to compare findings. The multi-engine mode is overkill for most work but makes sense for security-critical code where you want maximum coverage. The ADR enforcement is the real value here, it stops architectural drift by forcing you to either document decisions or classify changes as trivial. In CI mode it auto-generates proposed ADRs from git history so the pipeline doesn't break.
npx skills add https://github.com/alinaqi/claude-bootstrap --skill code-review