Security audit
Security checks for vulnerabilities and agentic risk
This is a disclosed Chinese technical blog-writing helper with minor privacy and scoping considerations, not evidence of malicious behavior.
Install only if you want an agent to help with Chinese technical blog writing and are comfortable providing a blog URL for style and SEO analysis. Prefer giving a public blog or pasted sample posts, and avoid sharing private drafts or sensitive pages unless you intend them to influence future writing.
The file states the skill is for writing Chinese technical blogs, and the metadata further labels it as 'chinese', but there is no indication that language choice is optional or user-selected. This is a natural-language locale constraint that appears forced rather than offered as an opt-in.
The README describes broad, automatic triggering for very common writing requests such as drafting, revising tone, and checking style. In an agent ecosystem, this can cause the skill to activate in situations the user did not explicitly opt into, unexpectedly steering outputs, collecting a blog URL on first use, or applying hidden writing constraints across unrelated tasks.
The activation description lists broad scenarios such as “写博客”, “调整语气”, and “问'下一篇写什么'”, which are common everyday writing requests and do not clearly distinguish when this specific skill should activate versus general writing assistance. The file also does not provide exclusion conditions or negative examples to narrow invocation scope.
The skill instructs the agent to visit and analyze the user's blog, including style, content layout, and SEO patterns, without an explicit notice about what data will be accessed, how it will be used, or whether external fetching will occur. This can create privacy and consent issues, especially if the blog contains personal identifiers, drafts, analytics-linked pages, or other unexpectedly sensitive content.
The instruction to 'remember' and reuse analysis results across future interactions establishes persistent retention of user-derived profile information without defined limits, retention duration, or minimization rules. This can lead to unnecessary storage of behavioral/style profiles and possible cross-session privacy leakage, especially if the remembered analysis includes personal identity cues or unpublished content characteristics.
