What problem does it solve? AI video generators produce each clip independently, so characters drift in face, hair, and wardrobe between scenes, and prompts written ad hoc lack cinematic specificity. This Skill turns a reel idea into a scene-by-scene script plus a prompt pack where every prompt carries a verbatim character block, locked color grade, IMAX camera language, negative prompt, and seed to hold consistency across scenes. ## Core Features & Use Cases - Consistency-locked prompt packs: Builds one rich 150-250 word prompt per scene via scene-prompts.mjs, self-verifying that the character block, grade token, IMAX token, and world token appear word-for-word in every scene. - Tool-agnostic workflow: The same prompts paste into Google Flow / Veo 3.1, Kling, Luma, Runway, Hailuo, Vidu, Pika, and PixVerse, with per-tool instructions for reference-image upload, negative-prompt fields, and seeds. - Two-stage audit harness: audit-reels.mjs runs automated checks (token consistency, prompt length, bridge flags, cinematic language), then a fresh reel-auditor subagent scores reel-worthiness out of 50 before delivery. - Use Case: A creator wants a 24-second cinematic action reel with the same character in four scenes. The Skill produces a character sheet with reference-image prompts, a scene script following the hook-agitate-payoff-loop arc, and a copy-paste prompt pack with a locked seed and negative prompts. ## Quick Start Ask the agent to create a cinematic reel script with scene-by-scene Veo prompts for your topic, keeping one character consistent across all scenes.
