The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the "markitdown" package from PyPI and cloning the source code from Microsoft's official GitHub repository.
[COMMAND_EXECUTION]: The documentation describes the usage of the "markitdown" CLI, package installation commands, and the execution of a Python script ("scripts/generate_schematic.py") for diagram generation.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (PDFs, Word documents, YouTube transcripts, etc.) to convert them into text for the agent, creating a vulnerability surface where malicious content in those files could influence agent behavior.
Ingestion points: Data enters the context through file conversion methods ("convert", "convert_stream") and URL-based conversion (YouTube).
Boundary markers: The instructions lack specific delimiters or instructions for the agent to ignore embedded commands within the processed content.
Capability inventory: The skill is configured with "Read", "Write", "Edit", and "Bash" tool permissions, which could be exploited if an injection is successful.
Sanitization: No specific sanitization or filtering of the converted text for malicious instructions is documented.
[DYNAMIC_EXECUTION]: The skill describes a plugin system that allows the loading of third-party extensions and encourages the discovery of these plugins on GitHub via hashtags, which involves running unverified external code.
[PROMPT_INJECTION]: The instructions include directives for the agent to adopt a specific persona ("Nano Banana Pro") and perform automated generation and refinement tasks, which attempts to override or define the agent's base behavior.

markitdown