shot-by-shot-analysis

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

• Third-party content exposure detected (high risk: 0.75). The required workflow ingests and analyzes “input video” sources (including their audible speech/text and visible text) and then writes those observations into the LLM-generated Markdown report, so any outsider-authored video content can become readable free text in the agent’s context via the video-analysis step.