regulatory-compliance

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: Indirect prompt injection surface detected.\n
  • Ingestion points: Untrusted content is ingested from files specified by the dpaPath parameter during agreement validation.\n
  • Boundary markers: The skill does not implement delimiters or instructions to ignore embedded commands in the processed documents.\n
  • Capability inventory: The agent has access to Read, Write, Glob, and Grep tools and executes local scripts.\n
  • Sanitization: No sanitization or validation is applied to the content of ingested files before they are processed by the agent.\n- [COMMAND_EXECUTION]: Execution of local scripts and hooks.\n
  • Evidence: The skill executes scripts/main.cjs to generate report templates.\n
  • Evidence: Pre- and post-execution hooks (hooks/pre-execute.cjs, hooks/post-execute.cjs) are utilized for validation and observability, involving the dynamic loading of local modules.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 11:45 AM