The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a bundled Python script scripts/render_docx.py to execute system commands (soffice, pdftoppm) via the subprocess module for document conversion and rendering.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted document content, creating a surface for potential instruction injection. Ingestion points: The skill reads .docx files through scripts/render_docx.py and python-docx as described in SKILL.md. Boundary markers: No specific delimiters or safety instructions for untrusted content are present. Capability inventory: The environment allows command execution via subprocess, as well as file system modifications via Write and Edit tools. Sanitization: There is no evidence of sanitization or filtering of document content before processing.
[PRIVILEGE_ESCALATION]: Documentation in SKILL.md includes commands for installing dependencies using sudo, which requests elevated permissions.

openai-doc