guard

Prevent destructive commands by enabling safety features that warn before executing risky operations like rm -rf, DROP TABLE, or force-push. This restricts edits to a user-specified directory and blocks actions that could delete data or modify infrastructure.

Restrict file edits to a specific directory boundary by applying a freeze boundary. This restricts Bash, Edit, or Write actions to a user-specified path, ensuring operations stay confined to your designated project directory during maintenance or automation.

Yes, you can get warnings before executing commands that modify production infrastructure. Destructive command warnings intercept risky operations, prompting confirmation before any action that could delete data or alter infrastructure is executed across development and production tasks.

The best way to protect production environments during debugging is to enable full safety mode. This combines destructive command warnings with directory-scoped edit protection, applying pre-tool hooks that verify boundaries before any Bash, Edit, or Write action executes.

Pre-tool hooks verify boundaries by intercepting Bash, Edit, or Write actions and checking them against a user-specified directory path. If the action targets a path outside the freeze boundary or matches a destructive pattern, the hook blocks it and issues a warning.

No, you do not need to manually configure sibling skills individually to enable directory-boundary edit protection. The guard mode integrates directly with sibling careful and freeze skills, automatically enabling destructive-command warnings and directory-scoped edits through a single setup.