Design Eye

The Problem It Addresses

Design prompts often fail when a theme like “parking” or “screw cap” is interpreted literally, rendered as a plain product shot, or filled with decorative elements. Design Eye turns the pre-generation design step into a controllable workflow: it derives structural concepts from the theme, matches a style system and master-level parameters, then produces prompts for image-generation models. Its focus is on why a visual should be composed in a certain way, not merely listing adjectives.

How It Works

• First-principles translation: decomposes input into literal, conceptual, emotional, cultural, and visualizable dimensions, then applies techniques such as substitution, scale inversion, absence-as-presence, and material contrast.
• Style and master reasoning: determines primary and secondary styles using style_matrix.md, then extracts style signature, composition rules, color logic, and design decisions from master_pool.md rather than naming masters without parameters.
• E-commerce hard rules: when the target is an e-commerce cover, it enforces banner ratios, contextual scenes, information hierarchy, and poster or cover terminology, while blocking white backgrounds, isolated products, and studio main-image views.
• Prompt generation: fuses main visual, copy, information layer, typography, color, and layout into a structured description, outputs Chinese and English prompts, and adds exclusion rules; if the request contains “you decide” or “fill this in” signals, it generates copy aligned to the chosen master’s tone.

Boundaries

The skill produces design prompts and structured judgments, not image files. Time-sensitive events, factual data, or holiday dates may trigger fact-checking; image attachments, out-of-library masters, or random mode depend on the referenced rules. It is best for early design reasoning in posters, covers, and brand visuals, and it does not replace final visual QA or commercial compliance review.