From short-video-director
Converts novel episode text into self-contained storyboard prompts for AI video models, ensuring continuity, pacing per config, and dialogue density self-checks (up to 3 rounds).
npx claudepluginhub wddxh/shortvideodirectorThis skill is limited to using the following tools:
- 从 `story/episodes/$ARGUMENTS[0]/outline.md` 的「本集资产清单」中提取本集引用的资产名称
Transforms short video scripts into self-contained storyboard prompts for AI video models, ensuring visual/audio continuity and state transitions with up to 3 self-check iterations.
Orchestrates story-to-video pipeline: breaks text into scenes, generates consistent Z-Image hero/refs + Qwen Edit frames, WAN FLF clips, ffmpeg concatenation.
Guides Next.js Cache Components and Partial Prerendering (PPR): 'use cache' directives, cacheLife(), cacheTag(), revalidateTag() for caching, invalidation, static/dynamic optimization. Auto-activates on cacheComponents: true.
Share bugs, ideas, or general feedback.
story/episodes/$ARGUMENTS[0]/outline.md 的「本集资产清单」中提取本集引用的资产名称assets/**/*.md 全部文件路径列表，仅读取文件名与清单中资产名称匹配的文件story/episodes/$ARGUMENTS[0]/novel.md — 必须读取story/episodes/$ARGUMENTS[0]/outline.md — 必须读取（含资产清单）config.md — 必须读取story/episodes/{上一集}/outline.md — 若上一集存在则读取story/episodes/{上一集}/storyboard.md — 若存在则读取末尾 2-3 个镜头skills/storyboarder-storyboard/rules.md — 必须读取并严格遵循（输出格式、字段约束、规则）$ARGUMENTS[0] — 当前集数（如 ep01）把本集小说原文转化为完整分镜——每个镜头是即梦视频模型的一次独立提交，提示词必须自包容（不能跨镜头引用）、画面与声音必须连贯叙事、状态必须与上下镜头衔接。下游消费者是即梦视频模型（按镜头独立生成视频片段）和 director-review-storyboard（审稿）。模型不会"理解整集剧情"——它只看本镜头提示词；任何跨镜头引用、画面与台词分离、抽象比喻都会让生成失败。
整体规划（动笔前）：
每个镜头的设计：
整集级自检：
skills/storyboarder-storyboard/rules.md — 必须读取并严格遵循生成分镜后，按照 skills/storyboarder-storyboard/rules.md 中的输出格式、字段约束和规则逐条自检（最多 3 轮）：
story/episodes/$ARGUMENTS[0]/storyboard.md