The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data from multiple formats (PDF, DOCX, HTML, YouTube transcripts) and injects the resulting text into the agent's context, creating a potential surface for indirect prompt injection. * Ingestion points: Conversion logic in scripts/batch_convert.py, scripts/convert_literature.py, and scripts/convert_with_ai.py. * Boundary markers: Absent; extracted content is directly appended to user/system instructions. * Capability inventory: The skill utilizes bash for execution and has Read/Write/Edit file permissions. * Sanitization: No sanitization or escaping is performed on the Markdown output before it is used by the agent.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the markitdown package from PyPI and cloning of source code from Microsoft's official GitHub repository. These sources are recognized as trusted and well-known.
[COMMAND_EXECUTION]: The SKILL.md instructions direct the agent to execute python scripts/generate_schematic.py for diagram creation, yet this specific script is missing from the provided skill files. This could cause the agent to attempt to execute non-existent local scripts or rely on external capabilities.
[REMOTE_CODE_EXECUTION]: The documentation describes a plugin system (--use-plugins) that enables the execution of third-party Python code. While this is a feature of the library, it introduces risk if the agent or user loads unverified plugins from external sources.

markitdown