werkvertrag-maengel

Pass

Audited by Gen Agent Trust Hub on Aug 6, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user data (legal fact patterns) without explicit boundary markers or sanitization instructions.
  • Ingestion points: User-provided inputs describing contract types, performance, and legal disputes in the 'Eingaben' section of SKILL.md.
  • Boundary markers: None identified in the provided instructions; inputs are interpolated into prompts without delimiters.
  • Capability inventory: The skill is restricted to legal reasoning and drafting; it has no access to dangerous system capabilities like shell execution, network writes, or file system modifications.
  • Sanitization: No evidence of input filtering, escaping, or validation of user-provided text.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 6, 2026, 10:47 PM
Security Audit — agent-trust-hub — werkvertrag-maengel