How to use the Japanese Seedance Examples agent skill

What is an agent skill, and what does this one do?

An agent skill is a folder containing a SKILL.md file. The file tells a compatible agent runtime when to use the skill and how to behave. This skill is triggered when you ask for Japanese Seedance 2.0 examples, Japanese prompt patterns, example rewrites, or safe versions of working Japanese video-generation prompts.

When it is active, the agent returns example prompts written in Japanese. Each example is labelled by risk: safe (original concept, no protected identity), needs-owned-reference (requires an asset you own, license, have public-domain rights to, or are authorized to use), surface-specific (depends on the web, API or workflow you use), or rewrite-required (mentions a protected identity, brand, celebrity, exact scene, song or voice).

• Keeps reference tags such as @Image1, @Image2, @Video1 and @Audio1 unchanged inside Japanese sentences.
• Rewrites protected names into original descriptions, for example turning a famous character into an original costumed character.
• Keeps emotion translated into framing, light, motion, sound or post-production notes.
• Returns the example, its label, a risk note and a safer Japanese variant when needed.

Install the skill from the ZIP download

Download the ZIP for skills/seedance-examples-ja from the page where you found this guide. Unzip it so that the skill folder, containing SKILL.md, sits in one of these locations: ~/.claude/skills/ for your user account, or .claude/skills/ inside a specific project.

If you use the Claude apps instead, open Settings, go to Capabilities, then Skills, and upload the skill there. The skill's files list contains only SKILL.md, so that is the file you need to keep in the folder.

• Personal install: place the unzipped folder in ~/.claude/skills/.
• Project install: place it in the project's .claude/skills/ folder so it applies only to that project.
• Claude apps: Settings → Capabilities → Skills, then upload.

Trigger the skill with natural requests

You do not need a special command. Ask in plain language for what you want, and mention Japanese and Seedance. The skill's description covers requests for examples, prompt patterns, rewrites and safe versions of Japanese prompts.

Good requests state the goal, the format and the constraints. Mention whether you need a product clip, a portrait, a dialogue line, or a textless social cut.

• Give me three safe Japanese Seedance 2.0 prompts for a product close-up with @Image1 as the product reference.
• Rewrite this Japanese prompt so it does not name a specific film or character, and keep @Image1 unchanged.
• Show a textless 9:16 Japanese prompt for a short social video; I will add subtitles in editing.

What the examples look like

The skill ships with six kinds of example in its instructions: product image-to-video, portrait micro-performance, an original sequence clip, a continuation of an approved clip, a line of dialogue, and a textless localization note. Each one describes camera, lighting and sound in Japanese, and keeps the subject original.

Examples are reference patterns to adapt, not finished prompts for every surface. Check each prompt against the video tool you actually use, since the skill itself does not generate video.

Requirements and safety notes

You need a Seedance 2.0 video-generation surface or workflow, which is not included in the skill, and a compatible agent runtime that loads SKILL.md files. The skill does not run scripts and does not use the network, according to its documentation.

The skill is not a Japanese language certification. Its own instructions say static validation does not check native-level quality and point to a separate multilingual language review step that is not included. Have a native speaker review important copy.

• Runs scripts: no.
• Uses network: no.
• Licence: MIT.
• Not for generating or reproducing real celebrities, branded characters or copyrighted scenes; the skill rewrites these instead.

Tips for better Japanese prompts

Keep subtitles, legal text and calls to action out of the generated frame and add them in editing, as the textless example suggests. Use the reference tags exactly as written, because the skill is designed to leave them unchanged.

When continuing a clip, state where the previous clip ended so the agent does not repeat the opening action.

• Translate feelings into visible things: a blink, a glance down, a soft light from one side.
• Name the camera framing and whether the camera is locked.
• Ask for a safer variant when a label is rewrite-required rather than forcing the original wording.