agentleFS
badgidsbadgids/Story-Film-Skills/skills/h3-prompt-writing/SKILL.md

h3-prompt-writing

Skill4 starsChanged 3 months agoApache-2.0View on GitHub

Write MiniMax H3 prompts for T2VA, I2VA, FL2VA, L2VA, and Ref2VA using the required audiovisual fields, keyframe alignment, speaker IDs, dialogue tags, reference labels, and timing.

11 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 10 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use stable <Subject N>, <Picture N>, <Video N>, and <Audio N> labels.

    line 45
  2. Very specific:

    Use <scenetrans> when verbal audio crosses a shot boundary and <cutoff> only when the requested clip actually truncates speech.

    line 52

“Common” is measured: kept by at least 6 in 10 of the 10 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

H3 Prompt Writing

This is Story-Film's portable implementation of the current public MiniMax H3 prompt-writing guidance. It is the mandatory formatting layer for every Story-Film MiniMax H3 prompt.

It contains no MiniMax Hub tool calls. Source behavior was reviewed at MiniMax-H3 commit d21241f0a4b3acbb34c97dae47fa417b7065e438; see SOURCES.md.

Workflow

  1. Identify the actual H3 mode supported by the selected workflow: T2VA, I2VA, FL2VA, L2VA, or Ref2VA.
  2. For T2VA/I2VA/FL2VA/L2VA read references/base-en.txt.
  3. For Ref2VA read references/ref-en.txt, and use the base guide for shared shot, camera, speaker, dialogue, and sound conventions.
  4. Preserve the required field names, section order, reference labels, shot numbering, and timing notation.
  5. Preserve user dialogue, lyrics, and visible scene text exactly in their original language.
  6. Hand the structured result back to minimax-h3 so Story-Film continuity, reference authority, exact-audio, and capture constraints are also enforced.

Base modes

Final core field order:

  1. integrated_multimodal_description
  2. overall_soundscape
  3. non_diegetic_music

I2VA, FL2VA, and L2VA also require the correct keyframe alignment instruction before the three fields. T2VA does not.

Ref2VA

Final section order:

  1. subject_definitions
  2. summary
  3. retention_analysis
  4. detailed_description
  5. overall_soundscape
  6. non_diegetic_music

Use stable <Subject N>, <Picture N>, <Video N>, and <Audio N> labels. Do not assign a reference a broader authority than the Story-Film source manifest allows.

Shared output rules

  • [Shot 1] has no cut timestamp. Later shots use strictly increasing At MM:SS.mmm cut times inside the requested duration.
  • Vocal sources use stable (S1), (S2), and later IDs.
  • Spoken or sung content uses <d>[Language] exact content</d>.
  • Use <scenetrans> when verbal audio crosses a shot boundary and <cutoff> only when the requested clip actually truncates speech.
  • Describe camera motion as natural shot action. Use motion type, amplitude, and speed only when they matter.
  • overall_soundscape contains ambience, physical action sounds, and non-verbal human sounds, not repeated dialogue.
  • non_diegetic_music describes audience-only score. Diegetic music belongs in the shot description.
  • Match the written audiovisual timeline to the requested H3 duration, normally 4 to 15 seconds per generated clip.

Done

The H3 format is exact, every reference label resolves, dialogue and visible text are preserved, and the final timeline matches the requested duration.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.