auskunft-rechnungslegung
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill requires the ingestion of external and potentially untrusted data which represents an attack surface for indirect prompt injection.
- Ingestion points: As described in
SKILL.mdunder the 'Sofortstart' section, the agent is instructed to collect original files, metadata, contracts, and platform-specific data (screenshots, logs, usage data). - Boundary markers: The instructions do not define clear delimiters for these external data points, nor do they instruct the agent to ignore any potential natural language instructions embedded within the ingested materials.
- Capability inventory: Across the skill instructions, the capabilities involve text processing and legal draft generation (memos, risk matrices, and claim plans). No tools for system command execution or network exfiltration were detected.
- Sanitization: The skill does not provide methods for sanitizing or validating the contents of the ingested files before processing them in the legal analysis workflow.
Audit Metadata