docx
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill follows secure development practices for AI agent extensions.
- XML Security: All XML parsing and manipulation utilize the defusedxml library in scripts/utilities.py and ooxml/scripts/pack.py, which effectively mitigates XML External Entity (XXE) and billion laughs attacks.
- Controlled Environment: The skill's main library (Document class in scripts/document.py) creates a temporary workspace for all operations, ensuring that original files are not modified until a successful validation step is completed.
- Validation Infrastructure: The skill includes a sophisticated validation suite (DOCXSchemaValidator, RedliningValidator) that checks for schema compliance and ensures that tracked changes follow the required protocols. It even uses soffice (LibreOffice) in ooxml/scripts/pack.py to verify that the final packed document is not corrupt by attempting a headless conversion to HTML.
- Command Execution: Use of subprocess.run in ooxml/scripts/pack.py and ooxml/scripts/validation/redlining.py is limited to necessary system utilities (soffice, git) with controlled arguments passed as lists, primarily for validation and generating diffs.
- Instructional Integrity: The provided documentation (SKILL.md, ooxml.md, docx-js.md) contains clear, professional guidelines without any signs of prompt injection or behavior-overriding patterns.
Audit Metadata