recoup-content-write-caption

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow reads artist voice from context/artist.md and context/audience.md (workspace-authored files) and only falls back to “real past captions” from /artists/{ROW_ID}/posts (outsider-authored public/social content) as few-shot anchors, which can include free-form caption text that the agent will ingest into the LLM context.