shots

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/extract-frames.py, the runtime ingests user-authored JSON at --shots-path/--manifest-path and then extracts frames based on shots[] fields (e.g., start_seconds, end_seconds, index) without selecting a specific trusted item first, so an outsider can submit a poison manifest for the LLM-adjacent workflow to read.