version: "1.0.0" name: style-brief-to-prompt description: Use when the user describes a look, atmosphere, or aesthetic direction in natural language and needs it converted into a clean English prompt fragment or style prefix for image or video generation.

Style Brief To Prompt

Use this skill to turn loose art direction into prompt language that generation models can actually use.

Trigger Cues

• "translate this style into a prompt"
• "turn this mood board into prompt language"
• "convert this Chinese style brief to English prompt"
• "give me a reusable style prefix"

Best For

• converting Chinese style notes into English prompts
• building reusable prompt prefixes
• turning mood boards into style language
• separating subject content from style language

When To Use

• The user mainly cares about look, tone, atmosphere, or art direction.
• The subject prompt already exists or will be written separately.
• You need a reusable style block that can be attached to many prompts.

When Not To Use

• The user needs full scene planning or storyboard logic.

Use creative-brief-to-storyboard.

• The user wants complete scene prompts from finished narration.

Use narration-to-visual-prompts.

• The user already supplied a final generation-ready prompt and only wants execution.

Workflow

1. Parse the style brief into concrete dimensions:

• palette
• lighting
• mood
• composition
• materials or texture
• era or cultural cues
• camera language if video is involved

1. If reference media exists, extract style signals first:

• ImageUnderstandingAgent with mode="style" or mode="prompt"
• VideoUnderstandingExpert with mode="prompt" if the reference is a video

1. Convert the style description into:

• one concise English prompt prefix
• one optional negative prompt
• one short summary of how the style should be applied

1. Keep the style prompt reusable. It should not hard-code the subject unless the user asked for subject and style together.
2. If the user wants multiple directions, return clearly separated variants instead of one overloaded prompt.

Handoff Examples

For extracting style from a reference image:

{"input_path":"inbox/reference.png","mode":"style"}

For combining the style prefix with image generation:

{"prompt":"<subject prompt>, <prompt_prefix_en>","provider":"gpt_image","size":"1536x1024","quality":"medium"}

For combining the style prefix with video generation:

{"prompt":"<subject motion prompt>, <prompt_prefix_en>","provider":"veo","mode":"prompt","aspect_ratio":"16:9","resolution":"720p"}

Output Contract

Return a structure like this:

{
  "style_summary": "...",
  "prompt_prefix_en": "...",
  "negative_prompt_en": "...",
  "applicable_to": ["image", "video"],
  "notes": "..."
}

Guardrails

• Prefer attributes, materials, eras, and camera language over direct imitation of living artists.
• Avoid copyrighted character or franchise mimicry when the user only wants a general aesthetic.
• Keep the prompt compact enough to combine with subject prompts later.
• Do not mix incompatible style cues unless the user explicitly wants contrast.
• If the user asks for several directions, keep each variant internally coherent rather than averaging them together.