markitdown
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides documentation and examples for using the 'markitdown' library, which is a reputable tool from Microsoft (github.com/microsoft/markitdown).
- [COMMAND_EXECUTION]: The skill includes command-line examples for installing and running the tool (e.g., 'pip install', 'markitdown file.pdf'). These are standard developer operations for the intended purpose of the skill.
- [EXTERNAL_DOWNLOADS]: The skill references official repositories and packages from trusted sources (Microsoft, PyPI). It provides instructions for optional cloud-based converters (Azure), which are well-known services.
- [INDIRECT_PROMPT_INJECTION]: The skill handles untrusted user data (PDFs, DOCX, URLs) for conversion. While this presents a theoretical surface for indirect prompt injection, the skill includes explicit security guidelines warning users to minimize capabilities and restrict paths when processing untrusted input.
- [DATA_EXPOSURE]: The skill mentions handling local files and URLs. It explicitly warns about audio privacy when using certain transcription features and highlights that cloud-based features (Azure) involve data transmission to those services.
Audit Metadata