safety-guard

You can prevent destructive file operations by activating Freeze Mode, which restricts agent file edits to a specific directory tree. This ensures the AI stays within the intended project scope while maintaining read access to the broader environment.

You can restrict file system write requests by activating Freeze Mode, which locks agent scope to a specific directory tree. This ensures the AI stays within the intended project scope while maintaining read access to the broader environment.

Guard Mode combines Careful Mode and Freeze Mode to provide maximum safety during autonomous agent sessions. It intercepts dangerous commands and restricts file edits to a specific directory tree simultaneously, ensuring comprehensive protection for production environments.

The best way to lock agent scope is using Freeze Mode, which restricts all file edit operations to a predefined directory tree. This prevents the AI from modifying files outside the intended project scope during autonomous execution.

This safety guard validates system commands and file operations against predefined safety patterns and directory constraints. It monitors shell commands and file system write requests to prevent destructive actions during autonomous agent execution in production environments.

Yes, you can combine both protections by activating Guard Mode. Guard Mode merges Careful Mode's command interception with Freeze Mode's directory restrictions, providing maximum safety during autonomous agent sessions in production environments.