The Agent Skills Directory

[SAFE]: The skill provides documentation and examples for using the markitdown library, which is an official tool from Microsoft. All external references target official GitHub repositories or well-known service providers.
[DATA_EXFILTRATION]: The skill includes explicit security warnings regarding Server-Side Request Forgery (SSRF) when processing remote URIs. It recommends that users manage network requests manually using standard libraries like requests before passing content to the conversion engine to ensure proper validation and control.
[COMMAND_EXECUTION]: The skill demonstrates standard command-line usage and Docker integration. These examples are consistent with professional development workflows and do not contain patterns for unauthorized command execution or privilege escalation.
[PROMPT_INJECTION]: The skill identifies the surface for indirect prompt injection inherent in document conversion workflows (e.g., processing untrusted PDFs). It frames this risk within the context of RAG (Retrieval-Augmented Generation) pipelines and focuses on structured extraction rather than executing content.
[REMOTE_CODE_EXECUTION]: The skill mentions a plugin system for extending capabilities. While dynamic loading is supported via standard Python entry points, the documentation emphasizes that plugins are disabled by default and requires explicit configuration by the user.

markitdown-awesome