The Agent Skills Directory

[COMMAND_EXECUTION]: The orchestration scripts scripts/convert.py and scripts/validate_output.py utilize subprocess.run to execute external command-line utilities such as pandoc, markitdown, and pdftotext. These calls use argument lists which mitigate shell injection risks, although they operate on file paths derived from user input.
[EXTERNAL_DOWNLOADS]: The skill instructions and scripts recommend using uv run --with or pip install to fetch well-known packages like pymupdf4llm, markitdown, and python-docx from the public PyPI registry. This is consistent with standard development practices for managing dependencies from trusted services.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection (Category 8) because it converts untrusted document files (PDF, DOCX) into Markdown content that the agent then processes as part of its context.
Ingestion points: scripts/convert.py and scripts/extract_pdf_images.py ingest and process external document files.
Boundary markers: The skill does not implement explicit boundary markers or 'ignore' instructions in the output Markdown to isolate document content from agent instructions.
Capability inventory: The skill environment has the capability to execute shell commands and trigger subsequent tool chains (e.g., /docs-cleaner).
Sanitization: Content is cleaned for formatting artifacts (e.g., fixing CJK spacing, removing Pandoc attributes), but there is no sanitization or filtering for natural language instructions embedded within the source documents.

doc-to-markdown