video-to-job-aid

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

• Third-party content exposure detected (high risk: 0.70). The required runtime workflow includes “Read its project” and “Transcribe and analyze the video's spoken and visual content” from a user-provided/uploaded training video and then pastes the confirmed job-aid text into the LLM reply; because the video/project content can be authored by someone other than the operating user, this can ingest outsider-authored free text (transcription/narration) into the agent context.