What problem does it solve? Teams need AI assistants to consistently route substantial, ambiguous design decisions through a governance pipeline, but manually editing CLAUDE.md is error-prone and hard to keep in sync. This Skill injects the governance policy as a single sentinel-wrapped block that is idempotent on re-run and cleanly removable. ## Core Features & Use Cases - Idempotent install and update: Appends the policy block to project or global CLAUDE.md, skips if already present, and replaces in place with --force. - Clean uninstall: Removes exactly the sentinel-wrapped block plus one adjacent blank line without deleting CLAUDE.md itself. - Approval gate with verification: Shows the exact bytes to be written, requires explicit user approval, then greps for exactly one sentinel to confirm the result. - Use Case: A team lead wants every Claude session in a repo to proactively run the govern pipeline on one-way-door decisions, so they run gov init to embed the policy once and re-run with --force after each plugin upgrade. ## Quick Start Ask the assistant to run gov init to install the governance policy into this project's CLAUDE.md.