What problem does it solve? AI-generated video often contains subtle defects like identity drift, flicker, pseudo-text, or watermarks that slip past casual review and reach production. This Skill provides a structured quality-control gate that inspects every candidate against the approved brief, storyboard, and rights report before post-production or publication. ## Core Features & Use Cases - Evidence-Based Inspection: Extracts metadata with ffprobe, pulls opening/closing/storyboard keyframes, builds contact sheets, and tests desktop and mobile crops. - Blocking Defect Detection: Rejects candidates for subject drift, deformed anatomy, flicker, unintended cuts, lost text-safe zones, visible watermarks, or unsafe content. - Structured Review Output: Produces a YAML review result with blocking and non-blocking findings, technical measurements, and responsible-stage routing to the correct upstream role. - Use Case: After generating a looping product video, run this Skill to verify the loop seam, confirm no watermark appears, check the mobile crop preserves the text-safe zone, and route any timing defect back to the scriptwriter. ## Quick Start Review the candidate video file against the approved storyboard and rights report, then return a structured approval or rejection with findings.
