MiniMax H3 Skills - Setup Guide¶
Source: minimax-ai/minimax-h3 (15,683+ combined installs on skills.sh) Category: Video Generation / AI Content Quality Tier: 🟡 Unverified (no skills.sh Trust Hub / Socket / Snyk verdicts published)
MiniMax H3 is MiniMax's flagship video generation model (Hailuo). The official MiniMax org publishes this repo (8,630★, 599 forks, created Jul 30 2026) with nine bundled agent skills: one portable prompt-writing skill (h3-prompt-writing) and eight style-specific generators built for the MiniMax Hub canvas workflow.
Installation¶
# Portable prompt-writing skill (recommended - works in any agent harness)
npx skills add minimax-ai/minimax-h3 --skill h3-prompt-writing
# Alternative URL form used in the README
npx skills add https://github.com/MiniMax-AI/MiniMax-H3 --skill h3-prompt-writing
# MiniMax Hub canvas skills (require the MiniMax Hub runtime)
npx skills add minimax-ai/minimax-h3 --skill brand-promo-video-generator
npx skills add minimax-ai/minimax-h3 --skill music-video-subtitle-generator
npx skills add minimax-ai/minimax-h3 --skill handdrawn-live-video-generator
Prerequisites¶
| Requirement | Details |
|---|---|
| Nothing for h3-prompt-writing | Pure Markdown + local reference files - no API calls, no keys, no runtime |
| MiniMax account | platform.minimax.io for H3 API access (generation only) |
| MiniMax Hub | The 8 canvas skills target the Hub's hub_generate_video / hub_generate_image canvas workflow |
| Hermes Agent | Any version that reads SKILL.md and local files |
Capabilities¶
h3-prompt-writing (portable - 7,857 installs)¶
Agent Skills format (name/description/compatibility frontmatter). Explicitly portable: "no external API calls, MiniMax Hub tools, or proprietary runtime required." Ships two prompt guides under references/: base-en.txt (text/keyframe modes) and ref-en.txt (full-reference Ref2VA mode).
| Mode | What it does |
|---|---|
| T2VA | Builds the full audiovisual timeline from text |
| I2VA | Starts from a first frame, develops forward |
| FL2VA | Describes the continuous path between first and last frames |
| L2VA | Infers an opening that converges on the supplied last frame |
| Ref2VA | Full-reference rewrites with subject_definitions, summary, retention_analysis, detailed_description, overall_soundscape, non_diegetic_music |
Key fields: integrated_multimodal_description, overall_soundscape, non_diegetic_music - exact field names, section order, and timing notation matter.
MiniMax Hub canvas skills (platform-bound)¶
| Skill | Installs | Notes |
|---|---|---|
| brand-promo-video-generator | 2,629 | Hub canvas workflow |
| music-video-subtitle-generator | 2,609 | Hub canvas workflow |
| handdrawn-live-video-generator | 2,588 | Hub canvas workflow |
| minimalist-product-ad-generator, 3d-animation-short-generator, papercraft-stop-motion-explainer, paper-collage-explainer-generator, co-op-game-intro-generator | - | Hub-bound; per README "not portable to generic agent harnesses" |
CorpusIQ Use Cases¶
| Use Case | How |
|---|---|
| UGC video prompt engineering | Apply H3 T2VA/I2VA prompt structures when scripting CorpusIQ UGC videos - the field ordering and timing notation are reusable outside MiniMax |
| Brand promo video briefs | brand-promo-video-generator for CorpusIQ product-ad briefs (requires MiniMax Hub) |
| Prompt-pattern research | base-en.txt / ref-en.txt are portable multimodal prompt-structure references - feed into the CorpusIQ UGC pipeline regardless of model |
| Multimodal answer enrichment | H3's audio-visual timeline format (integrated_multimodal_description + overall_soundscape) is a reusable structure for CorpusIQ visual answers |
Troubleshooting¶
| Issue | Fix |
|---|---|
| Hub skills fail outside MiniMax Hub | Expected - the 8 canvas skills call hub_generate_video etc. Use only h3-prompt-writing in generic agent harnesses |
npx skills add lists 9 skills but installs none |
Install per-skill with --skill h3-prompt-writing |
| Missing license | The repo has no LICENSE file (GitHub API reports none) - internal review recommended before commercial reuse of the prompt guides |
Verification¶
# List what the bundle provides
npx skills add minimax-ai/minimax-h3 --list
# After install, confirm the portable skill is present
ls ~/.hermes/skills/h3-prompt-writing/SKILL.md
Security¶
No skills.sh security audits are published for this publisher (Trust Hub / Socket / Snyk verdicts all absent as of Sep 14, 2026).
| Audit | Status |
|---|---|
| Gen Agent Trust Hub | Not published |
| Socket | Not published |
| Snyk | Not published |
h3-prompt-writing itself performs no network calls and ships only local Markdown reference files. The Hub-bound skills require a MiniMax account and the Hub runtime. Review the repo before adopting the Hub workflow into production pipelines.
Related¶
← Skills Catalog | Marketplace → Powered by CorpusIQ
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.