read-youtube-feed

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

• Third-party content exposure detected (high risk: 0.85). Yes—scripts/get_updates.py fetches outsider-authored YouTube HTML/JSON at runtime (requests.get to https://www.youtube.com/@{handle}/videos and https://www.youtube.com/watch?v={video_id}), extracts free-form text like video descriptions (shortDescription) and then generates summaries from that text, which the agent later uses as LLM-visible content when presenting updates and when calling read-podcast-workflow.