obliegenheitsverletzung-vvg
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADS
Full Analysis
- [SAFE]: No malicious code, obfuscation, or unauthorized data access patterns were detected. The skill's behavior is consistent with its stated purpose of providing legal analysis based on German insurance law.
- [EXTERNAL_DOWNLOADS]: The skill contains references to official and well-known German legal resources, including 'gesetze-im-internet.de' and 'dejure.org'. These are used for referencing statutes and court rulings and do not involve remote code execution or untrusted downloads.
- [COMMAND_EXECUTION]: The test file references a local evaluation script ('../../../scripts/eval.py'). This is a standard part of the development and testing workflow within the repository environment and does not involve untrusted remote execution.
- [PROMPT_INJECTION]: The skill uses structured instructions for its sub-agents (researcher, drafter, reviewer) without attempting to bypass safety filters or override system constraints.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided insurance data and case facts. While this represents a standard attack surface for indirect prompt injection, the skill lacks capabilities (such as file writing or arbitrary network requests) that could be exploited through this vector.
Audit Metadata