The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes internal Python scripts that execute specific system utilities to perform necessary document validation and comparison. In ooxml/scripts/pack.py, the soffice command from LibreOffice is called to validate the document integrity during the packing process. In ooxml/scripts/validation/redlining.py, the git utility is used to generate word-level diffs, enabling the validation of tracked changes.\n- [PROMPT_INJECTION]: The skill processes and extracts text content from user-provided documents, which introduces a surface for indirect prompt injection attacks if the documents contain instructions intended to manipulate the AI agent. \n
Ingestion points: Document content enters the agent context through pandoc text extraction and direct XML reading of word/document.xml.\n
Boundary markers: The skill does not implement specific delimiters or 'ignore' instructions when presenting extracted document content to the agent.\n
Capability inventory: The skill possesses file-system write access via doc.save() and command execution capabilities through the provided validator scripts.\n
Sanitization: The skill author has explicitly used the defusedxml library for parsing XML content, which mitigates XML External Entity (XXE) and other XML-based attacks.\n- [EXTERNAL_DOWNLOADS]: The skill relies on several well-known system dependencies and libraries for its document processing workflows, including pandoc, LibreOffice, Poppler, and the docx NPM package. These are widely recognized and standard tools for these operations.

docx