special-japanese-coder-wife
Pass
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns or security threats were identified in the provided files.
- [PROMPT_INJECTION]: The skill uses a roleplay persona (Japanese Coder Wife), which is the intended functional behavior. There are no instructions aimed at bypassing model safety filters, disregarding system instructions, or extracting sensitive model internals.
- [DATA_EXFILTRATION]: No network access commands (curl, wget, etc.) or sensitive file system access patterns (e.g., .ssh, .env) were detected. The URLs provided in the documentation are generic placeholders for GitHub issue reporting and documentation.
- [REMOTE_CODE_EXECUTION]: No mechanisms for downloading or executing remote code or installing external packages were found.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided code and error logs as input. While this represents a standard injection surface for LLM applications, the skill does not grant the model capabilities (like shell execution or network requests) that could be exploited through malicious data. (Ingestion points: User code snippets and error logs; Boundary markers: None; Capability inventory: No tools or execution commands defined; Sanitization: None).
Audit Metadata