visual-keywords_skill

Overview

This skill generates dense, keyword-rich strings optimized for fuzzy search and visual recall. It focuses on searchability and matching rather than prose, producing compact shorthand keyword blocks for images, illustrations, UI screenshots, or type specimens. Use it to improve indexing, retrieval, and similarity search across visual assets.

How this skill works

Given a visual input and a selected type (Aesthetic/Image or Font), the skill analyzes subjects, style, color, composition, mood, and technical traits. It synthesizes those observations into a concentrated, connector-free keyword string tuned for fuzzy matching and recall. Outputs follow strict shorthand format and length guidance to maximize discoverability without acting as alt text.

When to use it

• Index large image libraries for fast fuzzy lookup
• Tag design assets for creative search and mood boards
• Create searchable metadata for UI screenshots or design systems
• Generate searchable descriptors for type specimens and font catalogs
• Improve recall in multimodal retrieval systems

Best practices

• Choose correct type: Aesthetic/Image for visuals, Font for type specimens
• Include clear observations on subject, style, color, composition, and mood
• Avoid full sentences, articles, and connectors—use terse tokens only
• Keep output dense and focused: ~100–150 words for images, ~150–300 words for fonts
• Preserve technical details (lighting, perspective, weight, terminals) for higher precision

Example use cases

• E-commerce: tag product photos with style, color, material, and composition keywords
• Design ops: index UI screenshots by layout, grid, spacing, and visual hierarchy
• Creative search: build mood-based collections (e.g., ‘moody neon cityscape noir’)
• Typeface library: label fonts by category, weight range, distinctive features, and usage contexts
• Archival retrieval: convert photographer or illustrator catalogs into fuzzy-searchable keyword strings