guard

Guard mode prevents accidental destructive commands by activating pre-tool hooks that warn users before executing dangerous shell commands like rm -rf or force-push. It runs cross-skill checks to intercept these actions during development and production safety workflows.

Edit boundaries are directory-scoped restrictions that prevent unintended file modifications outside a user-specified path. Guard mode enforces these boundaries to ensure accidental changes cannot occur elsewhere in codebases or configuration repositories.

Activate guard mode to enable directory-scoped edit restrictions for your configuration repository. You must specify the target directory to restrict edits, preventing accidental changes to critical files outside that boundary during local debugging.

Yes, you need the sibling careful and freeze skills installed to use guard mode. Guard implements its destructive-command warnings and edit-boundary enforcement via pre-tool hooks that require these dependencies to run cross-skill checks.

Yes, you can use guard mode to protect production configurations while debugging locally. It combines destructive-command warnings with directory-based edit protections specifically designed for teams safeguarding critical files in production environments.

Directory-based edit protections rely on pre-tool hooks and require the sibling careful and freeze skills to function. If these dependencies are missing or bypassed, guard mode cannot enforce its destructive-command warnings or edit boundaries.