office-docx

Pass

Audited by Gen Agent Trust Hub on Oct 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes external system utilities to perform its core functions.
  • Evidence: ooxml/scripts/pack.py and ooxml/scripts/validation/redlining.py use subprocess.run to call soffice (LibreOffice) and git.
  • Context: These tools are used for document conversion, validation, and content comparison (diffing) within the defined redlining workflow.
  • [INDIRECT_PROMPT_INJECTION]: The skill handles untrusted data from external Word documents, presenting a potential attack surface.
  • Ingestion points: ooxml/scripts/unpack.py and scripts/document.py process XML content from user-supplied .docx files.
  • Boundary markers: Relies on structural OpenXML schema boundaries.
  • Capability inventory: Executes shell commands via subprocess.run for document processing and validation (ooxml/scripts/pack.py, ooxml/scripts/validation/redlining.py).
  • Sanitization: The skill consistently employs the defusedxml library for XML parsing, which provides protection against XML External Entity (XXE) and billion laughs attacks.
  • [EXTERNAL_DOWNLOADS]: The skill requires several standard third-party tools and libraries for operation.
  • Evidence: SKILL.md identifies pandoc, docx (npm), libreoffice, poppler-utils, and defusedxml (pip) as dependencies.
  • Context: These are well-known, legitimate document processing tools used for their intended purposes.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 26, 2026, 12:12 PM