docx
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: Processes untrusted Word documents which may contain malicious instructions. Ingestion occurs via pandoc and XML unpacking. No boundary markers are defined in instructions to isolate untrusted content from agent logic. Capabilities include subprocess.run and file I/O. Sanitized with defusedxml to prevent XML-based attacks.\n- [COMMAND_EXECUTION]: Benign use of subprocess.run for soffice and git validation. Uses argument lists to prevent injection.\n- [EXTERNAL_DOWNLOADS]: Documented system dependencies (pandoc, libreoffice, poppler-utils) and npm packages (docx).
Audit Metadata