| name | reverse-prompt |
|---|---|
| description | Use when turning an image into a reusable prompt for image generation, visual design, or cinematic direction. |
| phase | 1 |
| status | available |
Use this skill to convert an image into a structured prompt that captures its visual character.
- "turn this image into a generation prompt"
- "give me a prompt to recreate this style"
- "cinematic prompt from this frame"
- "reverse-engineer a prompt for this art"
- "score the image quality" → use
image-quality-audit - "what objects are in it?" → use
visual-reasoning - "review this UI screenshot" → use
screenshot-ui-review
- Run visual-reasoning on the image
- Extract subject, environment, lighting, palette, composition
- Translate signals into prompt language
- Return single-string prompt and token list
image_path— local path to image file
prompt— complete prompt stringtokens— individual prompt components
- Subject / objects
- Environment and setting
- Palette description
- Lighting (inferred from brightness + contrast)
- Composition style
- Depth / focus treatment
- Mood and tone
- Claiming exact camera or lens data unless EXIF confirms it
- Overfitting prompt language to a single model provider
- Hallucinated details not visible in the image