llm-prompting-guide — Claude Code skill: video and music prompt format
llm-prompting-guide is a Claude Code skill with prompt format rules for MiniMax H3 and MiniMax Music 3. Consult it when drafting prompts for these models in ComfyUI or other tools.
Apache-2.0 ★ 173 updated 2026-10-24
npx skills add sammcj/agentic-coding --skill llm-prompting-guide -a claude-codeWhat llm-prompting-guide does
This skill documents labelled prompt formats for two generative models. MiniMax H3 accepts text, image, or video reference and outputs video with stereo audio; MiniMax Music 3 generates complete songs from a structured caption and tagged lyrics. Both models degrade significantly when prompt structure is guessed rather than followed exactly.
The skill includes mode tables for each model—T2VA, I2VA, FL2VA, L2VA, R2V for H3—and a full corpus of music styles and production terms supported by Music 3. Check the appropriate reference before writing any prompt, and verify that every genre term and production phrase is in the model's training corpus.
What it helps with
- draft a video-to-audio prompt for MiniMax H3
- generate a song in MiniMax Music 3 with custom lyrics
- verify whether a genre or style is in the model corpus
- structure a prompt correctly for a generative model
Example prompts for the agent
- Write a prompt for MiniMax H3: person with rifle, midnight film noir
- Check my style keywords against the MiniMax Music 3 corpus
- Show me the prompt structure for video with native audio
How to install llm-prompting-guide
1Skills CLI (Vercel)
npx skills add sammcj/agentic-coding --skill llm-prompting-guide -a claude-code2Manually: put the files in place
~/.claude/skills/llm-prompting-guide/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/llm-prompting-guide/
1Skills CLI (Vercel)
npx skills add sammcj/agentic-coding --skill llm-prompting-guide -a codex2Manually: put the files in place
.agents/skills/llm-prompting-guide/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add sammcj/agentic-coding --skill llm-prompting-guide -a cursor2Manually: put the files in place
.cursor/skills/llm-prompting-guide/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install sammcj/agentic-coding Skills/llm-prompting-guide/SKILL.md2Manually: put the files in place
.github/skills/llm-prompting-guide/SKILL.md1Skills CLI (Vercel)
npx skills add sammcj/agentic-coding --skill llm-prompting-guide -a gemini-cli2Platform CLI
gemini skills install https://github.com/sammcj/agentic-coding --path Skills/llm-prompting-guideManually: put the files in place
.devin/skills/llm-prompting-guide/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about llm-prompting-guide
How do I install llm-prompting-guide in Claude Code?
Run npx skills add sammcj/agentic-coding --skill llm-prompting-guide -a claude-code, then open ~/.claude/skills/llm-prompting-guide/SKILL.md.
What does llm-prompting-guide do?
It's a reference for the exact prompt formats expected by MiniMax H3 (video and audio) and MiniMax Music 3 (song generation). It shows how to structure text for each model and which styles it recognizes.
Which models does this skill cover?
It covers MiniMax H3 and MiniMax Music 3 in ComfyUI and other tools. Do not use for chat-assistant prompts or generative models not listed in the guide.
Source text
SKILL.md · Apache-2.0
--- name: llm-prompting-guide description: Prompt format rules for generative video and music models. Use when writing or reviewing a prompt for MiniMax H3 (text/image/reference-to-video with native audio) or MiniMax Music 3 (song generation from caption plus lyrics), in ComfyUI or elsewhere. Do NOT use for chat-assistant prompts, or for generative models not named here. --- # LLM Prompting Guide The models below were trained on prompts written in a specific labelled structure, and guessing at that structure degrades output. Read the full guide for the model in play before writing any prompt text; the bullet here is a router, not enough to draft from. - MiniMax H3 (video + native stereo audio; T2VA, I2VA, FL2VA, L2VA, R2V) -> [references/minimax-h3.md](references/minimax-h3.md) - MiniMax Music 3 (full songs from a structured caption + tagged lyrics) -> [references/minimax-music-3.md](references/minimax-music-3.md) - Music 3 only responds to styles its training corpus contains. Check every genre term and production phrase before writing: [references/minimax-music-3-corpus.md](references/minimax-music-3-corpus.md) Adding a model: one bullet above, plus the skills `references/<model>.md` opening with a table of contents and a mode-selection table. By default, unless instructed otherwise save your generated outputs to a file in the current working directory named something related to the request.
© sammcj · Apache-2.0 · sammcj/agentic-coding