What problem does it solve? Writing prompts for image and video generation models often produces inconsistent styles, drifting characters, and unverifiable reference bindings across shots. This Skill standardizes how a production-grade generation prompt is constructed so every shot reuses a frozen style block, declares typed reference anchors, and enforces sound-picture and negative constraints. ## Core Features & Use Cases - Fixed Prompt Structure: Builds every prompt in a strict five-section order: STYLE LOCK, reference anchor table, audio track, shot sequence, and negative constraints. - Typed Reference Anchors: Defines a controlled <reference> tag vocabulary with roles (pov, character, voice, style-ref, motion-ref, storyboard, and more) so every cited asset is machine-verifiable and hallucination anchors are rejected. - Multi-Shot Continuity Rules: Enforces one information change per shot, first/last frame contracts between shots, and explicit HARD CUT markers for coherent sequences. - Use Case: When directing an AI short film, use this Skill to translate confirmed shot intent and character assets into a submittable prompt where the protagonist's character image and voice reference are bound, the style is frozen verbatim, and each shot's end state matches the next shot's start state. ## Quick Start Write a generation prompt for the next three shots of my video using the frozen style template, binding the main character image and voice reference, with per-shot start and end states.
