riffrec-feedback-analysis

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

• Third-party content exposure detected (high risk: 0.75). The extensive-analysis runtime path ingests outsider-authored free text when the input is meeting notes (read_notes() reads the .md/.txt body) or when the input is a Riffrec zip whose extracted session.json/events.json/transcripts contain user-spoken text; that text is then written into analysis.md/review-prompt.md and later used as LLM context for downstream brainstorm (via the created tunan:req GitHub issue body).