The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's Python script scripts/docling_word_to_markdown.py uses subprocess.run to execute the uv and docling binaries. This is used to install the necessary tools and perform the document conversion.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the docling package from public registries using the uv tool install command. These are well-known and standard development tools.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data by reading .docx files. This creates a surface where malicious instructions could be embedded in the document to influence the agent when it later processes the generated Markdown output.
Ingestion points: The input_docx file path provided to scripts/docling_word_to_markdown.py.
Boundary markers: None identified in the output Markdown generation logic.
Capability inventory: The script has capabilities to read/write files and execute subprocesses (uv, docling).
Sanitization: The script focuses on path normalization and image reference updates; it does not explicitly sanitize the text content for prompt injection vectors.

docling-word