docx
Pass
Audited by Gen Agent Trust Hub on Jun 30, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill follows secure coding practices for document manipulation and XML processing.
- XML Security: The skill consistently uses the defusedxml library and its components (minidom, sax) to parse XML content, mitigating risks associated with XML External Entity (XXE) and other XML-based attacks.
- Safe Command Execution: External tools such as LibreOffice (soffice) and git are invoked using subprocess.run with list-based arguments in ooxml/scripts/pack.py and ooxml/scripts/validation/redlining.py. This approach avoids the use of a shell, preventing shell injection vulnerabilities.
- Data Handling: The skill operates on copies of data in temporary directories, maintaining the integrity of original user files and preventing accidental data loss during processing.
- Input Sanitization: The Python library uses html.escape for user-provided strings (like author names) when inserting them into XML attributes, preventing XML structure breakage or attribute injection.
- Trusted Dependencies: All external dependencies listed (Pandoc, LibreOffice, Poppler, docx NPM package, defusedxml Python package) are well-known, established tools and libraries from official repositories.
- Indirect Prompt Injection Analysis: Potential ingestion of untrusted data from user-provided Office documents via ooxml/scripts/unpack.py and scripts/utilities.py is handled through defensive measures. Ingestion points are localized, and the capability inventory is restricted to localized file operations and safe subprocess calls. The presence of defusedxml provides sanitization against technical XML attacks, while the restricted execution environment limits the risk of content-based behavior overrides.
Audit Metadata