What problem does it solve? In AI-driven short-drama production, text-to-image prompts for characters, scenes, and props often drift from the script and asset bible, causing inconsistent visuals across episodes. This Skill provides a rigorous review workflow that catches omissions, contradictions, and ambiguous wording before prompts are sent to image generation tools. ## Core Features & Use Cases - Item-by-Item Comparison: Cross-checks every character, scene, and prop prompt against the full-series asset bible, final-asset-bible.json, and asset-registry.json for coverage and consistency. - Mental Rehearsal & Scoring: Simulates text-to-image output for each prompt, then scores characters, scenes, and props on a 1-10 scale with pass thresholds (average >= 8, no single item below 6). - Structured Verdict Output: Produces scoring tables and prioritized fix lists (must-fix vs. suggested) while returning only a single-line Chinese status to the producer. - Use Case: After the art-design stage generates 人物提示词.md, 场景提示词.md, and 道具提示词.md for a drama project, run this review to verify all cross-age character variants exist, scene four-grid formats are correct, and no prompt allows off-model generations. ## Quick Start Ask the AI to review the stage-two art direction outputs by checking the character, scene, and prop prompt files against the asset bible and registry, then report whether they pass.
