video-pattern-reverse-engineer

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

• Third-party content exposure detected (high risk: 0.75). The skill’s workflow explicitly takes “example videos from YouTube, TikTok, Reels, Shorts, or user-provided video files” and analyzes them at runtime, which means it will ingest outsider-authored free-form content (video transcripts/metadata/scene descriptions extracted from third-party platforms or files) into the agent’s LLM context.