How to Find Prompt Templates for UI Design in GPT-Image2

GPT-Image2 stores reusable UI design prompt templates in docs/templates.md under the section ### UI与界面, with machine-readable versions available in data/style-library.json and accessible via the built-in gpt-image-2-style-library skill.

The freestylefly/awesome-gpt-image-2 repository organizes its prompt templates for UI design across both human-readable documentation and structured JSON indices. Whether you need to copy a quick fill-in-the-blank prompt or integrate template retrieval into an automated workflow, the codebase provides three distinct access patterns optimized for different use cases.

Locating UI Design Templates in the Documentation

The primary entry point for prompt templates for UI design in GPT-Image2 is the comprehensive markdown guide located at docs/templates.md. This file serves as the canonical reference for all reusable prompt structures within the ecosystem.

The Main Template Catalog

According to the repository source code, the UI-specific content resides in the ### UI与界面 section of docs/templates.md. This section begins at anchor tpl-ui and organizes interface generation prompts into three distinct formats designed for varying technical requirements.

The markdown file contains fill-in-the-blank templates that specify:

• Platform targets (iOS, Android, Web)
• Visual style descriptors (minimalist, tech, skeuomorphic)
• Layout structures (top navigation, two-column, card flow)
• Aspect ratio constraints (9:16, 16:9)

Three Template Formats Available

The documentation identifies three specific template types for UI generation:

1. Plain-text "常规模板" – A direct copy-paste format for immediate use in chat interfaces
2. JSON "进阶模板" – A structured object format consumed programmatically by agents
3. 截图生成模板 – Specialized templates optimized for social-media-style screenshot generation

Accessing Templates Programmatically

For developers building automated pipelines or custom agents, the repository provides machine-readable alternatives to the markdown documentation.

The Style Library JSON Index

The file data/style-library.json contains a serialized index of all templates available in the system. According to the skill documentation in agents/skills/gpt-image-2-style-library/references/style-library.md, this JSON file maps template types to their structured definitions.

Entries in the JSON index include metadata such as:

• type: Classification identifier (e.g., "UI Screenshot")
• platform: Supported operating systems or display contexts
• style: Visual theme parameters including color palettes and typography
• constraints: Technical requirements like fidelity levels and aspect ratios

Using the Built-in Skill

The repository implements a dedicated skill called gpt-image-2-style-library that abstracts JSON template retrieval. Agents can invoke this skill using XML-style tags in prompts to automatically populate template structures with runtime variables.

Practical Implementation Examples

Copy-and-Paste Plain Text Template

For immediate use without infrastructure dependencies, extract the plain-text template directly from docs/templates.md:

为[产品类型]生成一张[平台，如 iOS/Android/Web]界面图。
核心功能：[功能点A]、[功能点B]、[功能点C]。
视觉风格：[极简/科技/拟物]，主色[颜色]，强调色[颜色]。
布局：[顶部导航/双栏/卡片流]，信息层级清晰，留白充足。
输出：高保真UI截图，文字清晰可读，比例[9:16/16:9]。

Node.js JSON Template Retrieval

To access the structured template programmatically, parse the style library JSON as implemented in the source repository:

// Load the style-library JSON
import fs from 'node:fs';
const lib = JSON.parse(
  fs.readFileSync(
    './data/style-library.json',
    'utf8'
  )
);

// Find the UI-Screenshot template
const uiTemplate = lib.find(
  t => t.type === 'UI Screenshot'
);

console.log('JSON UI template:', uiTemplate);

This script returns the JSON representation of the UI template (corresponding to lines 24-46 of templates.md), including all default parameters and structural schema.

Skill-Based Template Invocation

For agents integrated with the GPT-Image2 ecosystem, invoke the built-in skill to handle template population automatically:

<skill name="gpt-image-2-style-library">
{
  "type": "UI Screenshot",
  "platform": "Web",
  "product": "Dashboard",
  "layout": "Two‑column with top navigation",
  "style": {
    "theme": "Light",
    "primary_color": "Blue",
    "typography": "Sans‑serif"
  },
  "content": {
    "header": "Analytics Overview",
    "cards": [
      {"title":"Sales", "data":"$1.2M"},
      {"title":"Visitors","data":"34K"}
    ]
  },
  "constraints": "High fidelity, readable text, 16:9 aspect ratio"
}
</skill>

The gpt-image-2-style-library skill processes this input against the schema defined in data/style-library.json and returns a fully rendered prompt compatible with GPT-Image2's generation API.

Summary

• Primary documentation for UI design templates resides in docs/templates.md under the ### UI与界面 section, offering both plain-text and JSON formats

• Machine-readable index is maintained in data/style-library.json, enabling programmatic access to template structures

• Built-in skill gpt-image-2-style-library (documented in agents/skills/gpt-image-2-style-library/references/style-library.md) provides an abstraction layer for template retrieval and population

• Three access modes support varying workflows: direct copy-paste, JSON parsing, and skill-based invocation

Frequently Asked Questions

Where are the UI design prompt templates physically located in the repository?

The templates are stored in docs/templates.md within the ### UI与界面 section. For programmatic access, the same data is serialized in data/style-library.json and referenced by the gpt-image-2-style-library skill documentation located at agents/skills/gpt-image-2-style-library/references/style-library.md.

What is the difference between the "常规模板" and "进阶模板" formats?

The "常规模板" (standard template) is a plain-text fill-in-the-blank format designed for direct copying into chat interfaces. The "进阶模板" (advanced template) is a JSON-structured object containing typed fields and schema definitions, intended for consumption by automated agents and the gpt-image-2-style-library skill.

How do I use the templates in an automated workflow?

Import and parse data/style-library.json using Node.js or any JSON-capable runtime. Filter for entries where type equals "UI Screenshot" to retrieve the specific template object. Alternatively, invoke the <skill name="gpt-image-2-style-library"> XML block with your parameters to have the system handle template selection and population automatically.

Can I modify the existing UI templates?

Yes. The repository structure allows direct modification of docs/templates.md for documentation updates and data/style-library.json for schema changes. When editing the JSON file, ensure you maintain the structure expected by the gpt-image-2-style-library skill to preserve compatibility with existing agent implementations.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}