doc-coauthoring

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

• Third-party content exposure detected (medium risk: 0.65). The workflow explicitly instructs the agent to “read” a user-provided existing doc and to “pull context via available integrations (Slack, Drive, etc.)” and then incorporate that text into the LLM context; if those integrations/documents include outsider-authored free text (e.g., Slack messages from others or a shared doc written by someone else), that content would be ingested at runtime.