legal-risk-assessment
Fail
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: HIGHCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The instructions in the 'Historical Punishment and Credit Risk Assessment' section state that because the model cannot log in automatically, users should provide their own account and password information for the national credit system. This solicitation of credentials within the chat session is a high-risk pattern that can lead to credential theft or accidental disclosure.
- [PROMPT_INJECTION]: The skill processes untrusted external data, creating an indirect prompt injection surface. Ingestion points: User-provided 'internal compliance management documents' and 'inspection records' as specified in the input format. Boundary markers: None provided in the instruction set to separate data from commands or to instruct the model to ignore instructions within inputs. Capability inventory: The skill performs deep logical analysis and generates structured risk reports, with potential network interaction implied in the workflow. Sanitization: No validation, filtering, or escaping is performed on the content of the uploaded files.
Recommendations
- AI detected serious security threats
Audit Metadata