interview-synthesis

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow ingests the user-provided interviews.json (outsider-authored customer Q/A transcripts) and renders their free-text answers into the LLM context/output via scripts/interview_synthesizer.py (e.g., extract_snippets() → render_markdown() includes s['text']), creating an indirect prompt-injection risk if those transcripts contain adversarial text.