What problem does it solve? AI coding assistants can execute destructive git commands like git push --force or git reset --hard without confirmation, risking irreversible loss of commits, branches, or uncommitted work. This Skill installs a guardrail that intercepts and blocks those commands before they run. ## Core Features & Use Cases - Pre-execution blocking: A PreToolUse hook intercepts Bash tool calls and rejects commands matching dangerous patterns such as git push, git reset --hard, git clean -f, git branch -D, and git checkout .. - Project or global scope: Install the hook into a single project's .claude/settings.json or globally in ~/.claude/settings.json for all projects. - Customizable blocklist: Edit the bundled shell script to add or remove blocked patterns to match your team's safety policy. - Use Case: A developer lets Claude Code refactor a repository autonomously but wants to guarantee it can never force-push to main or wipe uncommitted changes; the hook exits with code 2 and a BLOCKED message whenever such a command is attempted. ## Quick Start Ask Claude to set up git guardrails that block dangerous git commands like push and reset --hard for this project.
