luo-xiang-perspective
Pass
Audited by Gen Agent Trust Hub on Apr 8, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill defines a clear persona with specific behavioral rules (e.g., using the 'Zhang San' case study method) without attempting to override agent safety protocols or system prompts. It explicitly instructs the agent to maintain a scholarly and审慎 (prudent) attitude towards sensitive topics.
- [DATA_EXFILTRATION]: No sensitive file access or network operations targeting non-whitelisted domains were found. The skill correctly directs users to the official National Laws and Regulations Database (flk.npc.gov.cn) for legal lookups, which is a well-known and trusted service.
- [REMOTE_CODE_EXECUTION]: The skill does not include any remote code execution patterns, script downloads, or untrusted dependency installations. The installation instructions use a standard repository pattern.
- [OBFUSCATION]: Analysis of all files, including research data in the
references/directory, revealed no encoded strings, hidden characters, or deceptive text patterns. - [COMMAND_EXECUTION]: The skill does not request or perform any shell command execution or privilege escalation. It functions entirely as a context-providing persona.
Audit Metadata