Security audit
Security checks for vulnerabilities and agentic risk
This writing skill is mostly a disclosed content-assistance package, but it makes hiding AI authorship a default behavior for publishable social content.
Review this skill carefully before installing if you publish under rules that require AI disclosure. It appears technically simple and non-persistent, but it is designed to make AI-assisted drafts read as human-written, so use it only where that is honest and allowed by your platform or organization.
The description explicitly promotes 去AI痕迹技术 to make content appear more natural, which normalizes concealment of AI involvement without any user opt-in or policy basis. This is risky because it can facilitate deceptive authorship presentation, evasion of disclosure norms, or circumvention of platform transparency expectations.
The trigger metadata is broad enough to overlap with ordinary writing-help requests, which can cause the skill to activate outside a clearly intended scope. Overbroad activation increases the chance that users are routed into persuasion-oriented or AI-trace-removal behavior without explicit intent or informed consent.
The activation scenarios list what the skill can do but does not define clear limits, exclusions, or conditions for use. In context, this makes the skill more dangerous because it includes content optimization and AI-trace-removal functions that could be invoked in ambiguous situations without the user's clear opt-in.
The skill repeatedly instructs removal of AI traces as a normal operating mode, which reinforces deceptive use rather than simple editing assistance. In a content-generation skill, this context makes the issue more serious because the functionality is directly tied to producing publishable material that may mislead readers or platforms about how it was created.
The document explicitly instructs authors to remove signs of AI generation and states that public-account content 'must' conceal AI traces, creating a default policy of undisclosed AI impersonation. This can mislead readers, bypass platform or organizational transparency requirements, and normalize deceptive content practices across multiple publishing platforms supported by the skill.
No suspicious patterns detected.
