How to use the Korean Seedance Vocabulary 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 an compatible AI agent when to use the skill and what guidance to follow. This skill's description says it should be used when you ask for Korean Seedance 2.0 prompt wording, Korean cinematic vocabulary, or translation of camera, lighting, action, VFX, audio, and production terms into Korean.

Once active, the agent draws on a table of Korean terms grouped by function: camera, lighting, motion, audio, and constraints. It writes a compact prompt that covers subject, action, camera, light, sound, and a preservation constraint. It also keeps reference tags such as image labels exactly as written and tries to keep dialogue politeness level explicit.

• Camera terms, such as 느린 돌리 인 (slow dolly in), 로우 앵글 (low angle), and 클로즈업 (close-up)
• Lighting terms, such as 역광 (backlight), 부드러운 창문 빛 (soft window light), and 림 라이트 (rim light)
• Motion phrases, such as 천천히 돌아선다 (slowly turns around)
• Audio phrases, such as 조용한 환경음 (quiet ambient sound) and 음악 없음 (no music)
• A constraint line, such as 로고, 라벨, 형태를 정확히 유지한다 (keep logo, label, and shape exactly)

How do you install the skill?

You can install it in one of two ways. Download the ZIP from the skill's page on this site and unzip it into your personal skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder. Alternatively, upload the skill in the Claude apps under Settings → Capabilities → Skills.

After installing, confirm the folder contains SKILL.md. The skill's only listed file is SKILL.md.

• Option A: download the ZIP, unzip it, and place the seedance-vocab-ko folder in ~/.claude/skills/
• Option B: place it in .claude/skills/ inside a specific project
• Option C: upload it in the Claude apps under Settings → Capabilities → Skills

How do you trigger it with natural requests?

You do not need a special command. Ask for Korean prompt wording, Korean cinematic vocabulary, or a translation of production terms, and the agent can load the skill based on its description. Say what the shot shows, which language you want, and any reference image you are using.

Keep the reference tag exactly as you typed it. The skill preserves tags unchanged, and their spelling is independent of the prompt language.

• Ask for a clip directly in Korean, for example: write a Korean prompt for a slow dolly in on a ceramic cup under window light
• Ask for a translation of production terms: translate this shot note into Korean camera and lighting terms
• Ask for a bilingual prompt that gives Korean wording with an optional English gloss

What does a finished prompt look like?

The skill uses a compact pattern: a reference line, one motion or lighting change, one camera move, and one sound instruction. The pattern below shows the structure. Replace the bracketed parts with your own content and a real reference tag.

• Subject and reference: keep the identity, product shape, or logo fixed
• Change: one action, light, or camera change only
• Camera: one movement
• Sound: one audio instruction

What do you need before you start?

The skill needs Seedance 2.0 or another SKILL.md-compatible agent to use the output prompts. It does not run scripts and does not use the network, according to its listing. Any reference images you want the model to follow must be supplied by you and referenced with real tags.

• An agent that supports SKILL.md skills
• Seedance 2.0 or another compatible agent to consume the Korean prompts
• Your own reference assets, if you use reference tags

What are the safety notes and limits?

The skill runs no scripts and uses no network access, so it only supplies text guidance. The skill's own notes say its Korean wording is pending locale-specialist review, and the independent review file is empty. Treat it as working production wording, not as native-quality dialogue or localization.

The skill also states that its table is not a proven performance rule. It does not guarantee better results from the video model. Check the generated clip yourself.

• No scripts run and no network use is listed
• Korean wording is pending locale-specialist review
• Do not rely on it for native-quality dialogue or localization
• It does not guarantee better model results

Tips for better Korean prompts

Describe what the viewer can see or hear. The skill recommends replacing vague style words with observable light, framing, blocking, or timing. Keep one camera move per shot so the instruction stays clear.

If you include dialogue, state the relationship between speakers and the speech level you want. The skill offers register choices only where they change the brief.

• Replace vague adjectives with a specific light source or framing
• Keep one camera move and one sound instruction per prompt
• Give the speaker relationship when dialogue is included
• Ask for an English gloss when you need to check the translation