Security audit
Security checks for vulnerabilities and agentic risk
This markdown-only writing skill has no executable payload, but it broadly enables making AI-written text appear human and can add opinions or first-person voice without clear user guardrails.
Review this skill before installing if you work with academic, legal, compliance, journalistic, employment, or publication content. Use it only on text you are authorized to revise, and do not use it to hide AI assistance, bypass disclosure rules, or add opinions or personal claims that are not actually the author's.
The README presents the skill as removing signs of AI-generated writing but does not warn users that this transformation can be deceptive or violate platform, academic, workplace, or publication rules. In this context, the omission is meaningful because the core function is not just general editing; it can be used to obscure authorship signals and bypass human review expectations.
The usage text says to 'invoke directly when editing documents,' which is broad enough to let the skill activate in many writing contexts without clear scope limits. For a tool explicitly designed to make text sound less AI-generated, that ambiguity increases the chance it will be applied to policy-sensitive, academic, journalistic, or compliance-related content where concealment or misrepresentation matters.
The skill explicitly instructs the agent to add opinions, first-person perspective, humor, and personality even when the user only asked for text cleanup. That can cause the agent to fabricate subjective stance, impersonate human authorship, or alter authorial intent in ways the user did not request, which is especially risky in professional, academic, or compliance-sensitive writing.
No suspicious patterns detected.
