guard

To restrict file edits to a specific directory, you enter your desired directory path when starting guard mode. The Skill enforces a freeze boundary that blocks any write operations attempted outside that user-selected directory.

Destructive command detection covers risky patterns including rm -rf, unsafe git operations, and destructive SQL statements. It validates execution against these patterns using companion hook scripts before any command runs.

To set up directory-scoped edit blocking, run the guard mode command and input the target directory path. This restricts all write operations to that boundary, preventing accidental out-of-scope edits during production debugging.

Yes, enforcing edit boundary validation relies on companion careful and freeze hook scripts. These scripts execute dual protection workflow checks to ensure destructive-command warnings and directory boundary blocks run together successfully.

You should avoid using a freeze boundary when performing broad repository-wide refactoring or multi-directory updates, because the directory-scoped edit blocking intentionally prevents writing to any path outside the user-selected boundary.