docx
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill reads text and metadata from Word documents via scripts/docx_read.py, scripts/docx_comments.py, and scripts/docx_revisions.py, which can contain untrusted instructions. Document content is returned as raw text in JSON format without explicit sanitization or delimiters to prevent the agent from executing embedded instructions.
- [COMMAND_EXECUTION]: The test file tests/test_docx_skill.py uses subprocess.run to execute local Python scripts. The documentation also recommends using the soffice command for PDF generation.
- [EXTERNAL_DOWNLOADS]: The skill depends on the python-docx library, which is downloaded from PyPI during installation.
Audit Metadata