Guide · Agent skill
How to use the Seedance 2.0 Video Prompting agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The seedance-20 skill is a set of Markdown instructions that teaches an AI agent how to turn a rough video idea into a clear, compact prompt for the Seedance 2.0 AI video model. It is for creators who want a short clip from an idea, marketers and producers planning multi-shot video, and developers who call Seedance 2.0 through a provider. After this guide you will know what the skill does, how to install it, how to ask for it in plain words, and what to check before any paid generation.
Seedance 2.0 Video PromptingEmily2040/seedance-2.0 · MIT Download the skill Quick start
- 1Download the ZIP. Get the seedance-20 skill ZIP from the download link on this page.
- 2Unzip into a skills folder. Extract it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
- 3Or upload in the Claude apps. Go to Settings, then Capabilities, then Skills, and upload the ZIP there.
- 4Describe your idea. Ask in plain words for the clip you want and say how it should feel.
- 5Answer the short questions. Give the surface, duration and any reference images the agent asks for.
- 6Review the compact prompt. Check the draft prompt and ask for changes before any generation.
- 7Generate only with authorization. Generate on your own account only if you choose to, and within your spending limit.
What the seedance-20 skill makes the agent do
Once installed, the agent follows the skill's procedure when you ask for Seedance 2.0 video work. It helps write text, image, video and reference-based prompts, including dialogue, lip-sync, audio and first/last frame control. It also checks platform facts such as which provider offers the model, pricing and model IDs, and it flags intellectual property or likeness issues before anything is generated.
The skill covers many surfaces, including Dreamina, Jimeng, CapCut, Doubao, Volcengine/Ark, BytePlus, fal, and third-party routers such as OpenRouter and Replicate-style providers listed in its description. For simple single clips it takes a short path; for sequences, real-person or brand material, and platform questions it uses fuller checks and routes to smaller sub-skills.
- Turns a rough idea into a compact prompt of roughly 40 to 110 words for a single clip
- Flags IP, likeness and brand risks before generation
- Answers API, pricing and model-ID questions from its source notes
- Keeps a running record of decisions across a project
Install the seedance-20 skill
Download the skill ZIP from the download link on this page. Unzip it into your personal skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder if you want the skill to apply only to that project.
If you use the Claude apps instead, upload the ZIP under Settings, then Capabilities, then Skills. The skill is then available in that app without any folder setup.
- Personal install: ~/.claude/skills/
- Project install: the .claude/skills/ folder inside your project
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Check what you need before you start
The skill needs an AI agent that can read Markdown files and the linked sub-skill files. Its own instructions are plain text, so no extra software is installed by the skill itself.
You only need access to a Seedance 2.0 surface, such as Dreamina, Volcengine, fal or a third-party router, plus its account or API key, if you choose to generate video. Planning and prompt writing do not require either.
- An agent that can open Markdown and linked files
- A Seedance 2.0 account or API key, only if you generate video
- Your own spending limits set on the provider side
Ask for it with natural requests
You do not need special commands. Describe the outcome you want in ordinary words, and name the surface or task if you know it. The agent reads the request and decides which part of the skill applies.
If the request involves a real person, a brand or a copyrighted character, say so early. The skill will move to its safety check before planning.
- Make a 5-second clip of a fox crossing a snowy road at dusk, I want it to feel lonely
- Write a Seedance 2.0 prompt that starts from this image and ends on a closed door
- Give me a prompt with a short line of dialogue and lip-sync for a shopkeeper
- Which provider offers Seedance 2.0 through API, and what model ID should I use?
Work through an example from idea to prompt
Start with the feeling or the outcome, not camera jargon. The agent will ask only the questions it needs, such as the surface, the duration and whether you have reference images. For a first draft it will treat your request as one clip and will not ask about a longer series yet.
Review the draft. If you want more shots, ask to continue or extend the clip, and the agent will move to its sequence checks. Keep the previous decisions in view, because the skill is built to remember them within a project.
- Say what you want the viewer to feel
- Say which surface you plan to use, if known
- Share references and say what role each one plays
- Ask for a revision or a longer sequence when the first draft works
Safety notes: scripts and network
The installed skill runs no scripts and makes no network requests. It is a package of Markdown files. The package's security notes say that one development-only evaluator script contacts a model provider and reads a credential from the environment, and that this script is excluded from the installed payload.
The skill cannot enforce permissions itself. Your agent host controls file access, network access and credentials. Never paste a secret or API key into a prompt, and check the authorized provider, assets and spending limit before any paid generation.
- Installed skill: no scripts, no network
- Private vulnerability reports go through the repository's Security tab
- Generation depends on your agent's tools and your account limits
What people use it for
- Turning a rough idea into a 40 to 110 word prompt for a single short clip
- Planning multi-shot video with consistent characters and style
- Writing dialogue and lip-sync lines for a short scene
- Setting first and last frame control between two reference images
- Checking which provider offers Seedance 2.0 and which model ID to use in an API call
Frequently asked questions
How do I install the seedance-20 skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder. In the Claude apps, upload the ZIP under Settings, then Capabilities, then Skills.
What do I need to use the seedance-20 skill?
You need an AI agent that can read Markdown files and linked files. You only need a Seedance 2.0 account or API key if you choose to generate video.
Is the seedance-20 skill free?
The skill is released under the MIT licence, as stated on its page. Any costs come from the Seedance 2.0 provider you use to generate video, which this guide does not price.
Can the skill make Sora, Veo or Kling videos?
No. The skill is scoped to Seedance 2.0 only and excludes Sora, Veo, Kling and Runway's own Gen models. It also excludes image-only prompting.
Why does my agent not seem to use the skill?
The skill's own notes say that a host must follow the linked files or provide its own routing. Check that the ZIP unzipped to the correct folder or uploaded correctly, then ask the agent to open the skill file directly.
Does the skill generate videos for me?
Not by itself. It guides the agent to write prompts and check facts. Generation depends on the agent's tools, your account and your spending limits.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use seedance-20 if you want help turning an idea into a clear Seedance 2.0 prompt, or if you need checks on providers, pricing and model IDs before you build. Pick a different tool if you need Sora, Veo, Kling or Runway's own Gen models, or if you only want image prompting.
Download the skill