The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted data.
Ingestion points: The skill is designed to ingest and convert untrusted external files (PDF, DOCX, PPTX, XLSX, etc.) through the MarkItDown.convert() and MarkItDown.convert_stream() methods, as documented in references/api_reference.md and implemented in scripts/batch_convert.py and scripts/convert_literature.py.
Boundary markers: Converted text content is appended directly to the agent's context without any protective delimiters or "ignore embedded instructions" warnings.
Capability inventory: The skill's environment is configured with Bash, Write, Edit, and Read tools, providing a high-impact execution environment if the agent obeys instructions embedded within a processed document.
Sanitization: No sanitization, validation, or escaping of the converted Markdown content is performed before it is presented to the language model.
[COMMAND_EXECUTION]: Instructions to execute an unverified script.
The SKILL.md file explicitly instructs the agent to run python scripts/generate_schematic.py for visual enhancements. This script is not included in the provided skill files, making it an external and unverifiable execution call.
[REMOTE_CODE_EXECUTION]: Support for unverified third-party plugins.
The documentation in references/api_reference.md and SKILL.md promotes a plugin system that loads external Python code at runtime. It encourages users to find and use plugins via public social tags (#markitdown-plugin), which could lead to the execution of malicious code from untrusted sources.

alterlab-markitdown