All products

Define a visual style with AI

Turn a description or reference into a structured style entry, compare it with an existing library and review its colours, textures and composition before saving.

Define a visual style with AI by turning a loose reference into specific decisions about colour, texture, lettering and composition. The style-suggest skill compares a description or one reference image with the existing catalogue, then recommends a close match or drafts a new structured entry. It produces text, not an image.

Keep the required shared visual library with the skill. The official whole-repository plugin route retains those files; the reviewed shell installer copies a different style library, so verify the actual schema and index before starting. Your chosen model receives the description and reference; a multimodal API route also sends the image to that provider. Review the proposed entry and index changes before approving a save. Similarity scores express model judgement rather than a measured guarantee of originality.

What it does

  • Convert a loose description into structured fields.
  • Separate background treatment from decorative elements.
  • Describe lettering by genre rather than a font file.
  • Check existing entries before adding a close duplicate.
  • Review the proposed file and catalogue row together.

Install and use

  1. Prepare the documented native plugin and check that common/visual-prompt-library contains its catalogue and schema.
  2. For example: “Describe cream paper, leaf motifs and simple printed lettering for a gardening notebook cover; check for an existing match first.”
  3. Review the comparison and proposed fields, replacing vague wording with visible characteristics.
  4. Approve the new file and index entry before saving. Creating images with it remains a separate downstream task.

FAQ

Source

Made by Mikhail Savchenko. Original skill: style-suggest, published on GitHub under the MIT licence.

View on GitHub