Security audit
Security checks for vulnerabilities and agentic risk
This skill is a local AI policy monitoring helper that scans user-provided markdown notes and does not show hidden network use, persistence, or destructive behavior.
Before installing, note that this is not a live regulatory feed: it scans markdown files you provide. Invoke it explicitly and point it only at policy-note directories you intend the agent to read; verify important legal conclusions against official sources.
The document title and introductory instructions are entirely in Chinese and direct AI agents and maintainers to use this file as the operating guide, but there is no indication that language choice is optional. This can be a natural-language policy issue when a skill effectively forces a specific language or locale without opt-in or justification.
The file mixes English and Chinese, but substantial required licensing and attribution content is presented only in Chinese or primarily Chinese, including mandatory-use wording such as '务必并读' and compliance instructions at the end. This can impose a locale/language requirement on users without an explicit opt-in or alternative fully equivalent language version, which matches the language/locale policy violation category.
The skill advertises broad trigger phrases like '最近AI出了什么新规' and '帮我盯AI政策', which overlap with ordinary user requests and can cause the skill to activate without clear user intent to invoke this specific tool. Because the skill is described as auto-running a scanner script, ambiguous invocation increases the risk of unintended tool execution and context switching, even though declared permissions are limited in this file.
The activation condition '收到「盯一下政策」类指令' is highly ambiguous and could match many routine conversational requests unrelated to this skill. In an agent environment, such vague matching can lead to accidental invocation of the skill and unnecessary access to local resources or execution paths referenced by the skill.
The documented flow states that generic natural-language phrases automatically invoke policy_radar.py, creating a direct path from broad user wording to script execution. Even if the script is intended for benign compliance work, auto-invocation on common phrases reduces user control and can be abused or triggered accidentally in multi-tool agent settings.
The Chinese trigger examples include everyday phrasing such as '帮我盯 AI 政策', which is broad enough to match normal advisory requests rather than a deliberate tool invocation. This makes accidental activation more likely, especially because the skill claims the system will automatically call policy_radar.py based on such phrasing.
The skill advertises very broad natural-language triggers such as general requests to 'keep up with AI policy' or 'help me monitor regulations', which can match many ordinary user intents beyond a narrowly scoped skill invocation. This can cause over-triggering or unintended activation, leading the agent to route unrelated compliance or policy questions into this skill and potentially execute scanning behavior when the user did not explicitly request it.
