aktenauszug-erstellen
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external documents such as court files (PDF/Word), which inherently creates an attack surface for indirect prompt injection if those files contain malicious instructions. However, the risk is minimized by the skill's rigid six-step output structure and its focus on factual extraction. \n
- Ingestion points: Processes user-provided court documents and evidence as defined in the Voraussetzungen section. \n
- Boundary markers: The skill does not explicitly define delimiters for external data, but uses a structured template for the final output. \n
- Capability inventory: Primarily text processing, summarization, and table generation. No arbitrary code execution or network transmission of processed data is present. \n
- Sanitization: No explicit sanitization of input text is mentioned. \n- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to verify legal norms and case law by accessing official and well-known German legal research portals (e.g., gesetze-im-internet.de, dejure.org, openJur). These network operations are intended for reference verification and do not involve sending sensitive case data to these services.
Audit Metadata