domainrecht
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions are strictly limited to legal analysis and do not invoke shell commands, file system writes, or unauthorized network activity. All referenced URLs are to official German legislative sites (gesetze-im-internet.de) or established legal portals (dejure.org).
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-provided fact patterns regarding legal disputes.
- Ingestion points: Untrusted data enters the context via the
fact_patterninput described in the skill metadata and test file. - Boundary markers: The skill implements a structured output template (DOMAINRECHT summary) to guide the agent's response format.
- Capability inventory: No tool usage, code execution, or network exfiltration capabilities are present.
- Sanitization: While no explicit sanitization of input is performed, the skill's narrow scope and lack of system-level tools minimize the risk of malicious payload execution.
Audit Metadata