Security audit
Security checks for vulnerabilities and agentic risk
This skill is a privacy-focused writing rewrite aid with optional metrics-only verification and no evidence of hidden data access or unsafe execution.
Reasonable to install for local rewriting and style cleanup. Treat optional verification carefully: only send aggregate metrics, not original text, rewritten text, secrets, credentials, or private content. Be aware that the skill may activate broadly on general polishing requests.
The phrase 'Use this skill when text sounds generic, repetitive, over-explained, overly polished, or obviously AI-shaped' relies on subjective descriptors without explicit trigger phrases, scope boundaries, or exclusion examples. In a markdown skill description, this can cause ambiguous activation because many ordinary writing tasks could match these broad conditions.
Lines L14-L15 state that no private text was sent to any verifier and that verification, if used, covers metrics only. In this file, however, that guarantee is presented as a checklist assertion rather than a technical control, so the documentation conveys a stronger privacy claim than the file itself substantiates.
