skills/klotzkette/claude-fuer-deutsches-recht/product-liability-prompt-code-public/Gen Agent Trust Hub
product-liability-prompt-code-public
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is purely instructional and follows standard legal analysis workflows without any detected malicious patterns, command execution, or network exfiltration.
- [NO_CODE]: No executable scripts, shell commands, or external dependencies are included in the skill. The instructions rely entirely on natural language processing within the agent's context.
- [INDIRECT_PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection as it is designed to ingest and analyze untrusted external data.
- Ingestion points: The
SKILL.mdinstructions guide the agent to process documents such as "Sachverhalt" (case facts), "Dokumente und Beweismittel" (evidence), and "Vertragsurkunden" (contracts). - Boundary markers: No specific boundary markers or "ignore instructions" delimiters are provided for the external data.
- Capability inventory: The skill does not define any executable tools or scripts, limiting the potential impact of an injection to the reasoning output.
- Sanitization: No validation or sanitization of input data is mentioned in the instructions.
Audit Metadata