The Agent Skills Directory

[PROMPT_INJECTION] (LOW): Detected a surface for Indirect Prompt Injection (Category 8). The skill processes untrusted Office documents and extracts text content for structural validation logs.
Ingestion points: ooxml/scripts/unpack.py and ooxml/scripts/validation/docx.py (via xml_files found in the unpacked directory).
Boundary markers: Absent. The text extracted from XML nodes is printed directly to the console output.
Capability inventory: subprocess.run (executing LibreOffice), shutil.copytree, and recursive file write operations.
Sanitization: The skill uses defusedxml for XML parsing which prevents XML External Entity (XXE) attacks, but it does not sanitize or escape the natural language text extracted from documents before it is displayed to the agent.
[COMMAND_EXECUTION] (SAFE): The ooxml/scripts/pack.py script executes the soffice (LibreOffice) binary via subprocess.run. This is used exclusively for headless document validation (converting the output to HTML to ensure it isn't corrupt) and uses a static command structure with local file paths.

pptx