The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted document data through its core workflows, creating a surface for indirect prompt injection.
Ingestion points: ooxml/scripts/unpack.py (ZIP extraction) and scripts/utilities.py (XML parsing).
Boundary markers: None explicitly implemented within document content processing to isolate instructions from data.
Capability inventory: ooxml/scripts/pack.py and ooxml/scripts/validation/redlining.py use subprocess.run to call external utilities like soffice and git. The Document library in scripts/document.py performs file system writes and XML modifications.
Sanitization: The skill effectively mitigates XML-related risks (like XXE) by using the defusedxml library for all parsing tasks.
[COMMAND_EXECUTION]: The skill executes external commands (soffice for document validation and git for diffing) to support its primary document processing functions. These operations are performed on files within a temporary workspace and use secure subprocess.run calling patterns without shell interpolation.
[EXTERNAL_DOWNLOADS]: The skill documentation identifies standard, well-known dependencies from established registries, such as pandoc and the docx Node.js package. These references are used for document processing and do not include unverified or risky scripts.

docx