xlsx
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes the LibreOffice binary (soffice) via subprocess.run to convert and render Excel files. The commands use absolute paths resolved via Path.resolve() and specific flags like --headless to ensure controlled execution.- [EXTERNAL_DOWNLOADS]: The skill's scripts specify dependencies on well-known Python packages (openpyxl, defusedxml, pypdfium2, pillow) which are fetched from the Python Package Index (PyPI) during execution via uv.- [INDIRECT_PROMPT_INJECTION]: The skill processes externally provided workbook files which could contain malicious data or formulas.
- Ingestion points: Excel files are loaded in scripts/office.py and scripts/recalc.py.
- Boundary markers: No explicit markers are used to separate user data from instructions in the agent's context.
- Capability inventory: The agent has the ability to execute file conversions through LibreOffice and perform filesystem operations.
- Sanitization: The skill uses defusedxml to prevent XML-related attacks and includes defensive guidance in references/workbook-features.md to prevent formula injection by treating external strings as literal text.
Audit Metadata