git-guardrails-claude-code

You can block dangerous git commands in Claude Code by installing a PreToolUse hook that intercepts destructive operations like push --force, reset --hard, and clean before execution. The hook stops the command and provides messaging with safe alternatives.

The PreToolUse hook prevents destructive git operations including push --force, reset --hard, clean, and branch -D from executing within Claude Code. It intercepts these defined commands to protect your repository from unintended data loss.

Yes, you can configure git safety hooks globally or per-project in Claude Code. This flexible scoping allows you to apply consistent guards against destructive git commands across your entire environment or isolate them to specific repositories.

Yes, jq is required to set up git guardrails for Claude Code. The configurable hook script relies on jq to parse tool input data and evaluate whether the invoked git command matches the defined blocklist of destructive operations.

Claude Code intercepts git commands before execution using a PreToolUse hook. This mechanism injects a script that evaluates the proposed command against a blocklist, halting destructive operations like reset --hard and clean while providing safe alternative guidance.

The best way to prevent accidental data loss from automated git commands is implementing a PreToolUse hook that blocks destructive operations like push --force, reset --hard, and clean. This ensures automated tools cannot execute irreversible repository changes without explicit approval.