careful

You can prevent accidental data loss by using a PreToolUse hook that warns you before executing destructive Bash commands. It matches patterns like recursive deletes, database drops, and hard resets, returning a permission decision that asks you to proceed or cancel.

Destructive shell commands to safeguard include recursive deletes, SQL DROP and TRUNCATE operations, force-pushing, hard resets, Kubernetes deletions, and Docker pruning. Safeguarding these commands warns you before operations that could cause irreversible damage to your production environment.

Yes, you can safely delete common build artifacts without warnings. The command detection includes safety exceptions for directories like dist, .next, and node_modules, allowing routine build-artifact deletions to execute normally while still guarding against irreversible damage.

You set up guardrails by enabling a PreToolUse hook that extracts your Bash command from tool input and matches it against destructive patterns. When a match is found, it returns a warning and asks for your permission before the command can execute.

Yes, destructive command detection works with Kubernetes deletions and Docker pruning operations. The system matches these destructive patterns and prompts you with a warning, ensuring you can choose to proceed or cancel before any irreversible infrastructure damage occurs.

The best way to add warnings is to use a command guardrail that detects SQL DROP and TRUNCATE operations. When a destructive database pattern is matched, you receive a warning and can manually override the workflow by choosing to proceed or cancel the execution.