Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add unknowlei/minimax-h3-opencode-skills --skill minimax-h3-reference-video-promptgit clone --depth 1 https://github.com/unknowlei/minimax-h3-opencode-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/unknowlei/minimax-h3-opencode-skills/minimax-h3-reference-video-prompt)<a href="https://agentmods.dev/skills/unknowlei/minimax-h3-opencode-skills/minimax-h3-reference-video-prompt"><img src="https://agentmods.dev/badge/skills/unknowlei/minimax-h3-opencode-skills/minimax-h3-reference-video-prompt/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/unknowlei/minimax-h3-opencode-skills/minimax-h3-reference-video-prompt"><img src="https://agentmods.dev/badge/skills/unknowlei/minimax-h3-opencode-skills/minimax-h3-reference-video-prompt.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 48 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00089 | $0.02413 |
| Opus 5 | $0.00044 | $0.01207 |
| Sonnet 5 | $0.00018 | $0.00483 |
| Haiku 4.5 | $0.00009 | $0.00241 |
Grade A, and why
minimax-h3-reference-video-prompt scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 146 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MiniMax H3 Full-Reference Video Prompt
Convert mixed reference assets and user intent into a traceable six-section full-reference prompt. Make every asset role explicit and write the target video as a detailed audiovisual timeline.
Routing Contract
New MiniMax H3 requests should enter through minimax-h3-creative-director. This is the default specialist for any request containing images. Route away only when the user explicitly declares pure first/last boundary roles and no image also carries identity, character, person, object, costume, scene, style, action, camera, composition, or another reusable reference responsibility. Ambiguity stays here. Full-reference may use prompt-level first/key/last-frame anchoring while maintaining consistency.
Official Format Authority
Before drafting, read ../h3-prompt-writing/SKILL.md and ../h3-prompt-writing/references/ref-en.txt. Consult ../h3-prompt-writing/references/base-en.txt for the shared shot, camera, dialogue, visible-text, and sound rules when needed. Treat these official MiniMax files as the canonical prompt-format specification. If this skill or its local references conflict with them, follow the official files.
Read references/reference-rules.md before drafting.
Confirmed Multishot Handoff
When the director supplies a confirmed multishot_plan, treat its shot count, timing, content, framing, performance, camera, transitions, sound, active references, and continuity decisions as already answered. Do not ask those questions again. Map every confirmed shot into detailed_description, keep reference labels and responsibilities stable, and carry the continuity ledger into retention and preservation instructions. Reopen a decision only when the plan is physically impossible, conflicts with a source asset, or exceeds the effective duration.
Official Input Envelope
For the official online product/API, keep output within 4-15 seconds at 24 FPS and the prompt within 7000 characters. Accept at most 9 images, 3 videos, 3 audio files, and 12 mixed files total. Each video or audio file must be 2-15 seconds; combined video duration and combined audio duration must each be at most 15 seconds. Audio cannot be the only reference type. Respect the documented per-file and API-body limits. Treat local ComfyUI constraints separately when they are narrower.
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago First seen · 146 lines · 89 tokens per session scan A 910125ee7402
minimax-h3-reference-video-prompt is a skill published in the GitHub repository unknowlei/minimax-h3-opencode-skills (112 stars, last pushed 1mo ago), licensed MIT. It adds 89 tokens to every session and 2,413 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
seedance-prompt
This skill should be used when the user asks to write, improve, translate, compress, or debug a Seedance 2.0 video prompt; mentions T2V, I2V, V2V, R2V, camera direction, prompt quality, or provides reference assets for a production-ready prompt.
multimodal-llm
Vision, audio, video generation, and multimodal LLM integration patterns. Use when processing images, transcribing audio, generating speech, generating AI video (Kling v3, Sora 2, Veo 3.1 std/lite/fast, Runway Gen-4.5 via gen4turbo), or building multimodal AI pipelines.
bootstrap
Bootstrap a greenfield project with parent-owned interviews and bounded plan, revision, and apply actions.
deep-research
Autonomous deep research on codebases and technical topics with structured report output via map-reduce explorer architecture.
mermaid
Create, validate, and repair Mermaid.js diagrams. Use when generating flowcharts, sequence, class, ER, state, or Gantt diagrams, or any visualization.
teach-me
Turn a 'teach me X' request into a single interactive HTML lesson, rendered Arcade-first, by emitting a lesson data model and assembling it with rp1 tooling.