docx
Pass
Audited by Gen Agent Trust Hub on Jul 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill utilizes standard, well-documented Python libraries (python-docx, zipfile) to manipulate Word document structures within the local environment.
- [COMMAND_EXECUTION]: The skill includes instructions to use the soffice (LibreOffice) command-line tool for converting legacy binary .doc files and for exporting documents to PDF. These operations are performed in a local shell sandbox.
- [PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection by ingesting untrusted data from external Word documents. While this is an inherent characteristic of document processing tools, it is noted as a potential risk factor. Ingestion points: SKILL.md (code snippets for reading paragraph and table text). Boundary markers: Absent. Capability inventory: Shell access (sandbox: shell), subprocess execution (soffice), and local file system read/write. Sanitization: Absent; the skill does not include routines to filter extracted text content.
- [EXTERNAL_DOWNLOADS]: The skill explicitly prohibits network egress and the installation of external packages, directing the agent to use preinstalled tools and libraries only.
Audit Metadata