The Agent Skills Directory

[SAFE]: The skill is dedicated to behavior-preserving code refactoring. Its core logic focuses on improving code reuse, removing dead code, and enhancing clarity without altering the program's output or side effects.
[SAFE]: The skill implements a robust safety mechanism by requiring a 'green baseline' and running verification scripts (refactor-verify) after every single modification. This ensures that any unintended changes in behavior are immediately detected and undone.
[COMMAND_EXECUTION]: The skill utilizes local Node.js scripts such as checklist.mjs and orchestrate.mjs. These are internal components of the refactoring pipeline and are used for status checking and conformance auditing rather than executing untrusted commands.
[PROMPT_INJECTION]: The instructions do not contain any patterns typical of prompt injection or safety filter bypasses. The skill specifically instructs the agent to 'stay in its lane' and avoid fixing security or correctness issues, instead flagging them for a separate review process.
[DATA_EXFILTRATION]: There are no network-capable commands (like curl or wget to external domains) or access patterns targeting sensitive files (like credentials or configuration secrets) within the skill's scripts or documentation.

refactor-simplify