The Agent Skills Directory

[SAFE]: The skill wraps the official markitdown CLI from Microsoft. It performs input validation and uses safe shell-scripting practices (e.g., array-based argument passing) to prevent command injection.
[EXTERNAL_DOWNLOADS]: The skill documentation correctly identifies that certain optional features, such as audio transcription and OCR, may involve data transfer to external services (e.g., Google or Azure). It explicitly instructs the agent to verify user authorization before proceeding with these specific conversions.
[COMMAND_EXECUTION]: The skill executes local commands (markitdown, git, basename, wc) to process files and determine output paths. These operations are scoped to the user's files and the standard ~/.agents/output/ directory, following established patterns for agent skills.

markitdown