Security audit
Security checks for vulnerabilities and agentic risk
This is a Markdown-only code hallucination review guide with no install steps, executable code, persistence, or credential access.
Installers should expect a Chinese-language code review aid focused on hallucination detection. If your environment auto-selects skills, consider invoking it explicitly only when you want hallucination checks, because its trigger terms overlap with ordinary code review requests.
The skill metadata and examples strongly bias the interaction toward Chinese output without indicating language should follow user preference. While not directly enabling code execution or data exfiltration, it can override expected response-language behavior and reduce usability, clarity, or operator awareness in multilingual environments.
The activation phrase set includes broad terms like '检测幻觉 / 验证代码 / 代码审查', which can overlap with ordinary user requests and cause the skill to trigger unexpectedly. That increases the chance of unintentional behavior injection into unrelated conversations, especially in environments where skills are auto-selected from natural language prompts.
No suspicious patterns detected.
