The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands like rg (ripgrep) and fd to identify code smells and potential targets for refactoring across the workspace.
[COMMAND_EXECUTION]: It utilizes the bun runtime for executing tests (bun test) and type checking (bun run typecheck) to validate simplifications before they are committed.
[COMMAND_EXECUTION]: The skill instructs the agent to perform Git operations, including surgical conventional commits for each logical change.
[SAFE]: The skill incorporates extensive safety mechanisms, including a mandatory pause and surface ritual that requires the agent to present findings to the user before modifying any code.
[SAFE]: A 'never-touch' reference file explicitly identifies security-critical strings (HKDF info labels, IndexedDB keys, Durable Object names) and public schemas that must not be changed, preventing accidental security regressions or data corruption during refactoring.

collapse-pass