legal-document-draft

Pass

Audited by Gen Agent Trust Hub on Jun 17, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes internal Python scripts (load_profile.py and grade_draft.py) to validate configuration files and the final drafted documents. This is a standard part of the tool's workflow for quality assurance and data integrity.
  • [EXTERNAL_DOWNLOADS]: No external code or binary is downloaded during runtime. The skill contains references to the official Taiwan 'Laws & Regulations Database' (law.moj.gov.tw) for statute citations, which is a well-known government service.
  • [PROMPT_INJECTION]: The skill implements defensive measures against model drift or training data interference. The grade_draft.py script specifically scans for and rejects content that incorrectly references non-Taiwanese legal frameworks (such as GDPR 72-hour notification rules), effectively enforcing the defined 'Path A' legal scope through structural and content-based validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 17, 2026, 05:05 PM
Security Audit — agent-trust-hub — legal-document-draft