The Agent Skills Directory

[SAFE]: The skill's primary purpose is auditing existing agent surfaces using local evidence such as git history, session logs, and configuration files.
[SAFE]: Contains clear scoping rules and 'failure shields' that prevent the agent from overreaching or self-auditing without explicit user permission.
[SAFE]: Accesses only relevant application-specific paths (like ~/.codex/plugins/cache) and strictly treats them as read-only evidence, explicitly forbidding modifications to cache files.
[SAFE]: No remote code execution, external downloads, or data exfiltration patterns were detected across any of the analyzed files.
[SAFE]: The skill utilizes standard, low-privilege diagnostic tools like git log and rg for analysis within the local project context.

skill-audit