Skip to content

T8mars/minimax-h3-prompt-skill-T8 / skills/write-seedance-video-prompts

write-seedance-video-prompts

Verified

Write model-native Seedance 2.0 video prompts from text, images, video, or audio references. Use for text/image generation, multimodal reference, video editing, extension, track fill, or combined tasks that need stable subjects, explicit asset roles, coherent event order, camera, sound, visible text, and a 4–15 second feasibility check.

FileKindSize
SKILL.mdskill md2.9 KB
agents/openai.yamlconfig269 B
references/summary.mdmarkdown954 B
references/template.mdmarkdown2.1 KB

references/template.md

View on GitHub
# Seedance 2.0 prompt template

## Task matrix

- `text_image_generation`: create a new video from text and optional still/keyframe references.
- `multimodal_reference`: transfer a declared subject, scene, action, camera, effect, style, voice, music, or continuity property.
- `video_editing`: modify a named `视频N`; leave unspecified content unchanged.
- `video_extension`: continue a named `视频N` while preserving audiovisual and narrative continuity.
- `track_fill`: connect two or three named videos in declared playback order.
- `combined`: reference one asset while editing or extending another.

## Planning card

```text
Target: [task family], [4–15 seconds], [aspect ratio]
Subjects: [stable label + 2–3 identity anchors]
Assets: [图片N/视频N/音频N -> one narrow role each]
Event chain: [setup -> trigger -> visible change -> payoff]
Primary camera: [one coherent behavior per shot]
Sound/text: [only requested or evidenced layers]
```

## Simple output

Use one compact paragraph:

```text
[Stable subject binding]. [Initial composition and spatial state]. [Visible trigger and continuous action with speed/force/contact]. [Primary camera behavior]. [Lighting, color, and medium]. [Synchronized sound/dialogue/music/visible copy when requested]. [Compact constraints].
```

## Multi-shot output

```text
镜头1:[composition + subject action + spatial change + camera + necessary sound]
镜头2:[causal handoff + new visible information + coherent camera + necessary sound]
镜头3:[payoff/final state + readable hold + necessary sound]
约束:[only identity, geometry, medium, contact, continuity, or copy protections]
```

Do not attach numeric time ranges to the shots. Let causal and spatial changes determine their length.

## Final audit

- Duration is 4–15 seconds and action density is feasible.
- Shot labels are consecutive when used.
- Every subject and asset label resolves in both directions.
- One shot does not contain contradictory simultaneous camera commands.
- Dialogue language, visible copy, sound, and music are not invented.
- No H3 field name, label, timestamp, or retention syntax remains.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 323kUpdated Claude CodeCodexVerifiedInstall