docx

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The skill explicitly unpacks and reads arbitrary .docx files (e.g., "pandoc --track-changes=all path-to-file.docx" and python ooxml/scripts/unpack.py <input.docx>) and mentions handling "someone else's document", so the agent will ingest and interpret user-supplied/untrusted document content.