office-custom

Pass

Audited by Gen Agent Trust Hub on Jul 9, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The script soffice.py acts as a wrapper for the LibreOffice command-line interface. It searches for the soffice or libreoffice binary in standard system paths (macOS, Linux, and Windows) and executes it using subprocess.run() to perform document conversions. Arguments are forwarded to the binary, and it uses a dedicated temporary directory for the user profile to ensure compatibility in restricted environments.
  • [COMMAND_EXECUTION]: The pack.py script optionally executes the skill's own validate.py script via subprocess.run() after creating an OOXML file. This is an internal check to ensure the generated ZIP archive and XML structure are well-formed.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external Office documents, which are untrusted data sources.
  • Ingestion points: unpack.py reads ZIP-compressed OOXML files; pack.py and lint_docx.py read XML and ZIP data.
  • Boundary markers: None. The scripts process the files structurally.
  • Capability inventory: File system write access (to the output directory/file) and subprocess execution (LibreOffice and internal validation).
  • Sanitization: The scripts use standard XML parsers (xml.etree.ElementTree and xml.dom.minidom). While these process the document content, they do not execute instructions contained within the text of the documents.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 9, 2026, 03:00 AM
Security Audit — agent-trust-hub — office-custom