grundstueckskaufvertrag

Pass

Audited by Gen Agent Trust Hub on Aug 3, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is designed for legal automation and contains no instructions that attempt to override AI safety filters or exfiltrate sensitive information.
  • [EXTERNAL_DOWNLOADS]: The skill links to official German legislation and judicial databases (gesetze-im-internet.de, dejure.org) which are trusted and well-known services for legal information.
  • [COMMAND_EXECUTION]: The skill invokes a local script (scripts.legal_calc.cli) for mathematical calculations related to legal costs and deadlines. This execution is limited to pre-defined legal parameters within the agent's environment.
  • [PROMPT_INJECTION]: The skill processes user-supplied contract details, which is a surface for indirect prompt injection. However, the instructions mandate verification of claims against primary legal sources, and the agent has no access to sensitive system resources. Ingestion points: user inputs for property details and contract terms in SKILL.md. Boundary markers: not present. Capability inventory: local command execution (scripts.legal_calc.cli) and task delegation to sub-agents. Sanitization: not explicitly specified.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 3, 2026, 08:18 PM
Security Audit — agent-trust-hub — grundstueckskaufvertrag