writer
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external and potentially untrusted data sources including ODT, DOCX, and CSV files.
- Ingestion points: Processes external documents for conversion and content manipulation; reads data from CSV/Spreadsheets for mail merge operations.
- Boundary markers: The skill suggests using "approved inputs" and disabling macros/updates, but does not provide technical enforcement of these boundaries within the instructions.
- Capability inventory: File system write access (
doc.storeToURL), shell command execution (soffice), and network listener configuration (uno:socket). - Sanitization: The instructions mention that manual XML replacement was removed due to lack of escaping, showing awareness of sanitization, but relies on the underlying LibreOffice engine for safety.
- [COMMAND_EXECUTION]: The skill utilizes the
sofficebinary via shell commands to perform document opening and headless conversions. - Evidence:
soffice --headless --convert-to docx document.odtand usage of shell loops for batch processing. - [EXTERNAL_DOWNLOADS]: The instructions prompt for the installation of external Python libraries to enable specific functionality.
- Evidence:
pip install odfpyandpip install ezodf. These are standard, well-known libraries for ODF manipulation.
Audit Metadata