Security audit
Security checks for vulnerabilities and agentic risk
This is a simple copywriting-style skill with no executable code or data access, though it strongly steers output toward Xiaohongshu-style Chinese copy.
Install this if you want a focused Xiaohongshu-style copywriting assistant. Be aware it may strongly bias responses toward that platform's Chinese copy style and mentions a persona file that is not included, but it does not introduce code execution, credential access, persistence, or broad local data access.
The skill description and persona constraints hard-code Chinese/Xiaohongshu-style output and require adherence to an external persona file without explicit user opt-in. This can override user preference, cause unexpected behavior, and create prompt-routing or policy conflicts, especially in multilingual environments or where output language/style should be user-controlled.
The invocation example uses a very generic natural-language phrase that could easily overlap with ordinary user conversation, increasing the chance the skill is triggered unintentionally. In an agent environment, broad triggers can cause unauthorized capability activation or route benign requests into this skill without clear user intent.
No suspicious patterns detected.
