miturheber-band-splits
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including chat logs, studio files, and contract documents provided by the user to determine co-authorship. This presents a theoretical surface for indirect prompt injection where instructions hidden within evidence could attempt to influence the agent's reasoning.
- Ingestion points: External documents, evidence files, and chat logs mentioned in the 'Arbeitsweg' and 'Beweise anfordern' sections.
- Boundary markers: None explicitly mentioned to separate user evidence from instructions.
- Capability inventory: The skill is limited to legal analysis and document generation; it does not possess capabilities for file writing, shell execution, or arbitrary network requests.
- Sanitization: Not specified for ingested documents.
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to perform live checks of legal norms and case law via official and established legal databases.
- Trusted sources: The skill references well-known official German and European portals such as gesetze-im-internet.de, dejure.org, openJur, and official court databases (BVerfG, BGH, EuGH).
- Context: These references are used for standard legal verification and do not involve executable code or untrusted scripts.
Audit Metadata