MiniMax H3 Ships Official Skills: Prompt Writing + 8 Style Generators
MiniMax released installable skills for H3 on GitHub: one h3-prompt-writing skill covering all five generation modes plus eight style-specific video generation skills with bilingual SKILL.md files.
MiniMax released an official skills bundle for MiniMax H3 in the model's GitHub repository (MiniMax-AI/MiniMax-H3/skills) on August 9-10, 2026. The bundle contains one prompt-writing skill and eight style-specific video generation skills, each packaged as an installable SKILL.md (with a Chinese SKILL.cn.md counterpart for the style skills) plus any reference materials it needs.
h3-prompt-writing
The core skill rewrites multimodal requests into H3's structured prompt format for all five generation modes: T2VA, I2VA, FL2VA, L2VA, and Ref2VA. It composes the three core fields (integrated_multimodal_description, overall_soundscape, non_diegetic_music), aligns keyframes, and defines reference labels for images, videos, and audio.
It ships with two prompt guides under references/:
base-en.txt— base text/keyframe modes (T2VA / I2VA / FL2VA / L2VA), covering shot structure, camera motion vocabulary (type + amplitude + speed), speaker IDs(S1), dialogue<d>blocks, on-screen text, and theoverall_soundscape/non_diegetic_musicsections, with four complete example promptsref-en.txt— full-reference (Ref2VA) mode, defining the six-section rewrite format (subject_definitions,summary,retention_analysis,detailed_description,overall_soundscape,non_diegetic_music) with<Subject N>/<Picture N>/<Video N>/<Audio N>label rules and a complete example
The eight style skills
| Skill | What it generates |
|---|---|
| 3d-animation-short-generator | Complete stylized 3D animated shorts: story outline, character/environment cards, shot planning, storyboards, generation, assembly, BGM |
| brand-promo-video-generator | Brand promo videos from product images and briefs |
| co-op-game-intro-generator | Co-op game intro sequences |
| handdrawn-live-video-generator | Hand-drawn / live-video hybrid explainers |
| minimalist-product-ad-generator | Clean minimalist product ad shorts with beat-synced typography |
| music-video-subtitle-generator | Music videos with subtitles |
| paper-collage-explainer-generator | Paper-collage style explainer videos |
| papercraft-stop-motion-explainer | Papercraft stop-motion explainers with character and scene continuity |
Each style skill walks through an ordered production workflow (brief → outline → cards → shot planning → generation → assembly → review), keeps character consistency and scene continuity, and documents what it is not for (e.g. no KOC talking-head ads for the product-ad skill).
Installation
Install with the skills CLI:
# List all skills available in this repository
npx skills add https://github.com/MiniMax-AI/MiniMax-H3 --list
# Install a single skill
npx skills add https://github.com/MiniMax-AI/MiniMax-H3 --skill h3-prompt-writing
# Install all skills
npx skills add https://github.com/MiniMax-AI/MiniMax-H3 --skill '*'The skills are agent-portable: h3-prompt-writing requires no external API calls, MiniMax Hub tools, or proprietary runtime, and the repository also mirrors it under .agents/skills/ and .claude/skills/ for agent-specific setups. The agents/openai.yaml file only adds optional ChatGPT/Codex UI metadata.
Availability
- Skills directory: MiniMax-AI/MiniMax-H3/skills
- Model page: MiniMax H3
- Related: MiniMax H3 open weights land in ComfyUI
Comments
Sign in with GitHub to join the discussion.