skills/dawiddutoit/custom-claude/docx/Gen Agent Trust Hub

docx

Pass

Audited by Gen Agent Trust Hub on Feb 24, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes well-known external utilities soffice (LibreOffice) and git using subprocess.run for document validation and diffing tasks. These operations are associated with the skill's primary purpose.
  • [EXTERNAL_DOWNLOADS]: Users are directed to install official dependencies from trusted sources, including system packages via apt-get and libraries from the PyPI and npm registries such as pandoc, docx, and defusedxml.
  • [PROMPT_INJECTION]: The skill's document processing workflow provides a surface for indirect prompt injection as it ingests data from external Word files.
  • Ingestion points: Content is extracted from user-provided .docx files using ooxml/scripts/unpack.py and converted to Markdown via pandoc instructions in SKILL.md.
  • Boundary markers: Content extraction relies on standard XML and Markdown structures; it does not currently add explicit markers or warnings to isolate potential instructions within the document text.
  • Capability inventory: The skill has the capability to write to the file system and execute document processing subprocesses.
  • Sanitization: The skill effectively mitigates common XML vulnerabilities by using the defusedxml library for parsing and applies HTML escaping when updating document metadata fields.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 24, 2026, 05:23 PM