Minimax H3 Creative Director
Act as a creative director for MiniMax H3 video. Turns a goal such as an ad, a recipe short, a game teaser, a music clip or an explainer into a shot list with a consistent cast, writes H3-ready prompts, prices the whole piece, renders drafts and finals, and iterates on feedback. Use when the user wants a finished multi-shot piece rather than a single clip, or asks how to get consistent characters or products across MiniMax H3 shots.
Install
npx skills add seraphinawolfe/minimax-h3-creative-agent-skill --skill minimax-h3-creative-directorgit clone --depth 1 https://github.com/seraphinawolfe/minimax-h3-creative-agent-skill.git /tmp/minimax-h3-creative-agent-skill
mkdir -p ~/.claude/skills
cp -r /tmp/minimax-h3-creative-agent-skill/. ~/.claude/skills/minimax-h3-creative-directorPersonal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.
From the SKILL.md
# MiniMax H3 creative director
You plan like a director and spend like a producer. The deliverable is a set of rendered
shots (and, if the user has ffmpeg, one cut). Everything is written to a brief file and run by
`scripts/brief.py` in this skill's folder (Claude Code: `${CLAUDE_SKILL_DIR}/scripts/brief.py`).
## 1. Intake: ask at most five questions
Ask only what you cannot infer, in one message, with defaults the user can accept by saying "go":
1. Where will it run? → aspect ratio (Reels/Shorts/TikTok 9:16, YouTube 16:9, feed 1:1 or 4:3, cinematic 21:9).
2. How long in total? → number of shots (each shot 4–15 s; 3 shots of 5 s is a 15 s spot).
3. What must look the same in every shot? → the cast: people, product, place, logo, voice, track.
4. Which of those exist as files? → images (JPEG/PNG), clips (MP4/MOV, 2–15 s), audio (MP3/WAV, 2–15 s).
5. Budget? Default: drafts first, finals only after a yes.
## 2. Plan the piece
- **One action per shot.** If a sentence needs "then", it is two shots.
- **Cast once, reuse everywhere.** Every recurring element is a cast member with one reference file, passed to each shot that shows it. This, not prompt wording, is what keeps a face or a label stable.
- **Pick each shot's mode:**
| Shot needs | Mode | Brief fields |
|---|---|---|
| a fresh scene | text | `prompt` |
| a specific person, product, place, motion or beat | reference | `uses: [cast names]`, mention them as `{name}` |Details
- Skill name
minimax-h3-creative-director- License
- MIT
- Repo updated
- 19 October 2026
FAQ
What does the Minimax H3 Creative Director skill do?
Act as a creative director for MiniMax H3 video. Turns a goal such as an ad, a recipe short, a game teaser, a music clip or an explainer into a shot list with a consistent cast, writes H3-ready prompts, prices the whole piece, renders drafts and finals, and iterates on feedback. Use when the user wants a finished multi-shot piece rather than a single clip, or asks how to get consistent characters or products across MiniMax H3 shots.
How do I install the Minimax H3 Creative Director skill?
skills CLI (Claude Code, Cursor, Codex and more): npx skills add seraphinawolfe/minimax-h3-creative-agent-skill --skill minimax-h3-creative-director. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.
Is Minimax H3 Creative Director popular?
It lives in seraphinawolfe/minimax-h3-creative-agent-skill, which has 37 GitHub stars and 0 forks.
Related skills
Video Storyboard Gen
26anyhui/video-storyboard-gen-skill
为中文视频创作生成可执行的故事结构、逐镜分镜表和即梦/Seedance 等 AI 视频提示词。用户提到分镜脚本、故事板、镜头设计、短视频/广告/宣传片脚本、运镜拆解、图生视频或文生视频提示词时使用；也适合把主题、文案或创意改造成带时间码的拍摄方案。不用于实际调用视频平台、规避平台规则，或编造产品功效与素材授权。
Korean Thriller Asset to Video
29Work-Fisher/fisher-storylens
将韩剧式悬疑、复仇小说或剧本按角色生图提示词确认、场景生图提示词确认、用户上传图片后生成MiniMax视频提示词的三阶段流程制作，支持修改造型与连续片段，默认中文交付。
Founder Voice Ghostwriter
22BayramAnnakov/founder-voice-ghostwriter
Collaborative ghostwriting for founders and experts. Extracts first-hand knowledge through structured interviews and transforms it into authentic, non-AI-sounding content. Use when user wants to write thought leadership, blog posts, founder stories, comparison pages, or says things like "help me write", "make this authentic", "rewrite in my voice", or mentions avoiding AI detection.
Douban Sync
33cosformula/douban-sync-skill
Export and sync Douban (豆瓣) book/movie/music/game collections to local CSV files (Obsidian-compatible). Use when the user wants to export their Douban reading/watching/listening/gaming history, set up incremental sync via RSS, or manage their Douban data locally.
Adhx
19itsmemeworks/adhx
Search your own ADHX saves (posts kept from X, Instagram, TikTok, YouTube, Reddit and more) by meaning, including what is in pictures and videos, and fetch any X/Twitter post as clean LLM-friendly JSON. Use when the user asks to find something they saved, bookmarked or watched ("find my saved videos with a frog in them", "that article I saved about sleep", "the TikTok with the guy building a boat"), or shares an x.com or twitter.com post link (or a link to a post on their ADHX) and wants to read, analyze or summarize it.
Read Anything
31heyuxuan0209/read-anything
即时分析：任意链接（网页/博客/YouTube/B站/小宇宙播客/X）或本地录音（会议录音/语音备忘录）→ 抓取或本地转写 → 中文结构化解读（精读稿/快扫/访谈拆解/会议纪要/论文拆解/选题素材六种模板+自定义）→ 基于材料的多轮问答。触发：用户丢链接/音频文件/粘贴纪要并说"分析/解读/精读/这是什么/讲了啥/值得看吗/提选题"，或明确说"即时分析"。
Browse the most popular Claude skills, all images, video & audio skills, or write your own with the skills generator.