Security audit
Security checks for vulnerabilities and agentic risk
This is a lightweight humor skill that generates self-roast ratings, with no code execution, data access, persistence, or hidden security-impacting behavior.
Install this only if you want playful self-roast humor. Avoid invoking it for serious mental-health, distress, or support conversations, and ask for a milder tone if the roast style feels too harsh.
The description advertises '毒舌' behavior but does not clearly warn that outputs may include insulting, harsh, or self-deprecating roast language. That omission can lead to users, including minors or emotionally vulnerable users, invoking the skill without informed consent and receiving unexpectedly degrading content.
The trigger conditions are broad and instruct immediate activation on common self-deprecating phrases, which can cause unsolicited skill takeover in ordinary conversations about mood, productivity, or mental state. In context, this increases the chance of the agent responding with roast-style content when the user did not explicitly request it, creating safety and user-trust risks, especially for vulnerable or distressed users.
The file states '中文输入 → 中文输出；英文输入 → 英文输出', which imposes a language behavior automatically rather than offering the user a choice. Under the policy, forcing a specific language or locale without opt-in can be a natural-language policy issue unless it is clearly justified or user-selectable.
No suspicious patterns detected.
