cra-vulnerability-obligations
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted data from an external MCP service, representing a potential indirect prompt injection surface. It mitigates this via Ground Rule 2, which explicitly defines boundary markers by instructing the agent to treat tool results strictly as data and ignore any embedded instructions. Rule 3 further provides sanitization by prohibiting the use of sensitive information in queries.
- [DATA_EXFILTRATION]: While the skill communicates with an external gateway (gateway.ansvar.eu), it includes comprehensive data protection measures. Ground Rule 3 strictly forbids the transmission of secrets, credentials, or personal data, restricting external communication to generic regulatory and technical terms.
- [EXTERNAL_DOWNLOADS]: The skill utilizes a documented MCP connector from the vendor (davila7/Ansvar Systems) to fetch official regulatory texts. Ground Rule 2 ensures citations are verified against official government and EU domains before being presented to the user.
Audit Metadata