The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/render_docx.py uses subprocess.run to execute system commands including soffice (LibreOffice) and pdftoppm (Poppler).
The commands are used to convert DOCX files to PDF and then to PNG for visual inspection.
Arguments are passed as a list, which is a secure method to prevent shell injection, and the operations are restricted to temporary directories and user-specified paths.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external .docx files, which represents an indirect prompt injection surface.
Ingestion points: The workflow in SKILL.md involves reading DOCX content for review and extraction.
Boundary markers: The skill does not define specific delimiters for interpolating document content into the agent's prompt, though it recommends visual rendering as a validation step.
Capability inventory: The skill includes capabilities to write files (python-docx), create directories, and execute rendering commands via subprocess (scripts/render_docx.py).
Sanitization: No explicit sanitization or filtering of document text is performed before processing.

doc