youtube-claims

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

• Third-party content exposure detected (high risk: 1.00). Runtime LLM context is populated from the fetched YouTube transcript JSON (scripts/fetch.py downloads captions from YouTube, then scripts/segments.py slices it and prompts/inventory_claims.md passes the sliced transcript text into the LLM); this is outsider-authored free text (the video creator’s captions/transcript).