docx-official

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

• Third-party content exposure detected (high risk: 0.75). The required “Redlining workflow” ingests the target DOCX’s text/markup at runtime via pandoc --track-changes=all ... -o current.md and then the agent uses that markdown to plan edits, so outsider-authored document content can enter the LLM context (indirect prompt injection risk).

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The skill explicitly instructs running privileged installation commands (e.g., "sudo apt-get install pandoc/libreoffice/poppler-utils"), which asks the agent to perform actions requiring sudo and thus modifies the machine's state.