clause
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates as a purely instructional and checklist-based tool for reviewing legal documents. It contains no executable scripts, external binary dependencies, or network-enabled commands.
- [PROMPT_INJECTION]: No evidence of direct prompt injection or attempts to bypass system safety filters was found. The skill includes explicit instructions to maintain professional boundaries, such as refusing to provide legal advice and verifying all cited statutes to prevent hallucinations.
- [DATA_EXFILTRATION]: There are no commands or patterns indicating the exfiltration of sensitive data. The skill explicitly prohibits logging personal information or confidential user content.
- [COMMAND_EXECUTION]: The skill does not contain any shell commands, subprocess invocations, or dynamic code execution patterns. It relies entirely on standard model output for reporting.
- [PROMPT_INJECTION]: While the skill processes untrusted user-provided legal documents (a surface for indirect prompt injection), the risk is mitigated by a strict multi-phase workflow (SCOPE → SCAN → ASSESS → REPORT → SUGGEST) and a mandatory requirement to assign risk levels and citations based on pre-defined reference checklists. This structured approach prevents the agent from treating document content as high-priority instructions.
Audit Metadata