youtube-outliers

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

• Third-party content exposure detected (high risk: 0.90). The scripts (scripts/scrape_youtube_outliers.py and scripts/update_transcripts.py) explicitly scrape YouTube via yt-dlp and an Apify actor ("karamelo/youtube-transcripts") to ingest user-generated transcripts which are then passed into the Anthropic summarization prompt, so untrusted third-party content is read and can directly influence model outputs and downstream actions.