Security audit
Security checks for vulnerabilities and agentic risk
This skill is a markdown-only writing aid for reviewing technical articles, with no hidden code execution, credential access, or persistence beyond normal skill installation.
Before installing, be aware that broad trigger phrases may make this skill engage on some general article-review prompts. Use explicit wording when you want it, and review any proposed article edits before applying them.
The post-processing trigger phrases are very broad and map ordinary review requests directly to a workflow that scans, rewrites, and verifies article content. In an agent environment, this can cause unintended activation on loosely related prompts, leading the skill to take over editing behavior without a clear user opt-in or scoped boundary for what content should be modified.
The activation phrases include short, generic triggers such as '角色标注', '审稿重心', 'narrative focus', and 'review narrative', which can plausibly match ordinary editing or article-review requests outside this skill’s intended scope. In an agent-skill system, overly broad activation can cause the skill to engage unexpectedly and steer responses or workflows when the user did not explicitly request this specialized analysis, creating prompt-scope confusion and unintended behavior.
The trigger phrase "review narrative" is broad enough to match ordinary writing/editing requests that are unrelated to this skill’s narrowly defined purpose. Over-broad activation can cause the wrong skill to intercept user requests, leading to unintended instruction injection into unrelated workflows and reducing reliability of agent behavior.
The language selector emphasizes the English README as the current/default view while offering Chinese as an alternate, but there is no stated user language preference mechanism. This can be interpreted as a locale/language default rather than an explicit opt-in choice.
This file is a Chinese README and all usage instructions are presented in Chinese, while the policy requires avoiding forced language or locale constraints unless the user is given a choice or the constraint is clearly justified. Although there is an English link, this document itself does not state that language is user-selectable for outputs or interaction behavior.
No suspicious patterns detected.
