Security audit
Security checks for vulnerabilities and agentic risk
The skill is a coherent writing editor, but its silent pre-send mode can broadly rewrite substantive replies without telling the user.
Install only if you specifically want an automatic prose-quality gate that may silently alter substantive agent replies. For safer use, limit it to explicit edit or detect requests, or require disclosure when pre-send changes are made.
The skill explicitly instructs the agent to 'silently inspect' and in pre-send mode to 'return only the corrected substantive reply,' which enables undisclosed modification of user-facing output. Hidden rewriting is dangerous because it removes transparency and user control, making it easier for the system to alter meaning, tone, or commitments without the user realizing edits occurred.
The skill defines pre-send mode for broad categories like 'interpretations, advice, research synthesis, consequential explanations, and other developed replies' without precise activation boundaries. That ambiguity can cause the agent to apply automatic editing to many normal responses, creating unpredictable behavior and increasing the chance that user-intended wording, nuance, or safety-critical phrasing is altered without explicit consent.
No suspicious patterns detected.
