The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill directs the installation of the 'markitdown' library and its associated extras from official sources using standard package managers. This is necessary for the skill's primary function of document conversion.
[COMMAND_EXECUTION]: Use of shell commands is required for setting up Python environments, installing dependencies, and running the provided batch conversion utilities. The included Python scripts incorporate security best practices, such as resolving paths to prevent traversal and skipping symbolic links by default.
[DATA_EXFILTRATION]: The documentation identifies features that involve sending data to external providers, such as Azure Document Intelligence for OCR and Google Web Speech for audio transcription. It provides clear instructions on obtaining user consent and managing credentials securely before using these optional services.
[INDIRECT_PROMPT_INJECTION]: As the skill processes untrusted document formats that might contain malicious instructions, the documentation provides extensive guidance on mitigating indirect prompt injection. This includes treating all converted Markdown as untrusted data and separating it from system instructions in downstream agent workflows.
[DYNAMIC_EXECUTION]: The skill describes a plugin system that allows for the execution of Python code. It correctly advises that plugins should be disabled by default and requires a thorough review of third-party code before enablement to prevent arbitrary code execution.

markitdown