The Agent Skills Directory

[SAFE]: The skill provides instructions for a code smell audit process using a set of 56 reference cards. The workflow includes a manual verification step where candidates are checked against these static cards, which serves as a control against incorrect findings.
[SAFE]: A validation script scripts/validate-cards.sh is provided to ensure the integrity and structure of the reference cards. Analysis of the script shows it performs local string processing and filesystem checks without dangerous operations or network access.
[SAFE]: The instructions for large codebase audits recommend fanning out sub-agents with limited context ('nothing more' than the target list and relevant index sections), which is a security-positive application of the principle of least privilege.
[SAFE]: No suspicious network operations, obfuscation, or persistence mechanisms were found. The skill operates on the local repository using standard git tools (git diff, git status) and writes reports to a local docs/audits/ directory.

code-smells-audit