urhebervertragsrecht-verguetung
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, which could potentially contain hidden instructions to influence the agent's behavior.
- Ingestion points: Processes user-supplied contracts, evidence, platform logs, and version metadata (SKILL.md).
- Boundary markers: The instructions lack explicit delimiters or specific warnings to the agent to treat content within processed documents as data rather than instructions.
- Capability inventory: The agent is tasked with generating legal memos, risk assessments, and correspondence based on the provided inputs.
- Sanitization: There are no defined mechanisms for sanitizing or filtering instructions that might be embedded in the provided legal documents.
- [EXTERNAL_DOWNLOADS]: The skill requires the agent to perform live verification of legal sources from external domains.
- Fetches and verifies legal text and case law from established German and European portals including gesetze-im-internet.de, dejure.org, openJur, and eur-lex.europa.eu.
Audit Metadata