The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands specifically for repository auditing and verification. This includes git ls-files and find to determine project scope, and user-defined test/lint commands (e.g., npm test, pytest) to verify that auto-fixes do not break functionality. All execution is gated by an initial 'Pre-flight consent gate' where the user must explicitly approve the operation.
[SAFE]: The skill follows security best practices for automated editing. It implements an 'assume-malicious' posture toward its own operations by strictly limiting 'Safe' fixes to local, non-exported symbols. It uses a verification loop to revert changes that cause test regressions and provides a detailed audit log of every modification made to the filesystem.

brooks-sweep