Security audit
Security checks for vulnerabilities and agentic risk
This is a Markdown-only prompt-engineering guidance skill with broad but disclosed activation language and no code, network access, persistence, or hidden data handling.
Install this if you want Chinese-language prompt-engineering help. Be aware it may activate for many prompt-related requests and tends to answer in Chinese, so users who need another language or very narrow routing may want to adjust those preferences.
The trigger list is very broad and includes common phrases about getting better AI answers, prompt writing, and output formatting. This can cause the skill to activate for routine requests outside its intended scope, leading to prompt hijacking of normal conversations and reducing routing precision.
The matching conditions rely on ambiguous high-level concepts such as 'AI提问', 'AI回答不好', and '控制输出格式' without strong activation boundaries. In practice this increases accidental invocation risk, allowing the skill to intercept broad categories of harmless user requests and override more appropriate specialized skills.
Forcing Chinese output regardless of user preference can cause instruction conflict, user confusion, and unreliable downstream behavior in multilingual environments. While not directly a security exploit, it can degrade control alignment and produce incorrect handling when another language is required for safety, compliance, or interoperability.
Line L846 recommends fixing multilingual issues by explicitly requiring the model to use Chinese or English throughout. This is a natural-language locale policy constraint that appears to prescribe a fixed language choice rather than offering the user a choice or documenting a justified locale limitation.
No suspicious patterns detected.
