manage-vietnam-legal
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements strong internal safety guardrails. It explicitly instructs the agent to mask sensitive personal data (PII) and confidential commercial terms, and requires qualified human legal counsel sign-off for all outputs via the AI-LEGAL-GUARDRAIL.
- [SAFE]: No external network operations or unauthorized code executions are performed. The skill only uses local file operations such as read_file and write_file within its defined scope.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external legal documents and contracts via the read_file and search_code tools as described in SKILL.md. It includes boundary markers in the form of instructions to reject prompt injection attempts in examined contracts (ASI01). Capabilities include write_file, edit_file, and create_file to output analysis results. Sanitization is addressed through instructions to redact sensitive commercial terms and mask restricted personal data.
Audit Metadata