aktenzeichen-und-vorgangsnummer
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill consists entirely of natural language instructions for the agent to follow when processing legal and administrative documents in a German context (e.g., BEG IV, OZG). It explicitly instructs the agent to verify information via official sources and avoid hallucinations.
- [NO_CODE]: There are no scripts, executables, or code blocks within the skill definition. The logic is handled through natural language instructions.
- [PROMPT_INJECTION]: There are no attempts to override agent behavior, bypass safety guidelines, or extract system prompts.
- [DATA_EXFILTRATION]: No network operations or commands that could exfiltrate data were found. The mention of external legal databases (e.g., dejure.org) is for manual or tool-assisted verification of legal norms, which is standard behavior.
- [INDIRECT_PROMPT_INJECTION]: While the skill is designed to process external documents, it does so through a structured workflow that emphasizes verification and caution, particularly for lay users, which minimizes the impact of potential adversarial content in processed documents.
Audit Metadata