how-to-guide

How to Integrate gpt-image-2-style-library with Claude Code and Codex: Complete Developer Guide

August 31, 2026 freestylefly/awesome-gpt-image-2 ↗

You can integrate gpt-image-2-style-library with Claude Code via a .claude-plugin/marketplace.json manifest and with Codex as an npm package by importing the generatePrompt function from @gpt-image-2/style-library.

The gpt-image-2-style-library is a reusable skill module in the freestylefly/awesome-gpt-image-2 repository that provides structured prompt templates for GPT-Image-2. Both Claude Code and Codex can consume this library using provider-specific integration patterns while sharing the same underlying data source in data/style-library.json.

Architecture Overview

The repository uses a single-source-of-truth design. All style definitions, categories, and templates live in one JSON file, while thin adapter layers handle Claude Code and Codex compatibility.

Component	Purpose	Location
Data layer	Master catalog of styles, scenes, and templates	[`data/style-library.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/data/style-library.json)
NPM package	Distributable skill with programmatic API	`agents/skills/gpt-image-2-style-library/`
OpenAI manifest	Example LLM tool definition	[`agents/skills/gpt-image-2-style-library/agents/openai.yaml`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/agents/skills/gpt-image-2-style-library/agents/openai.yaml)
Claude plugin	Claude Code marketplace registration	[`.claude-plugin/marketplace.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/.claude-plugin/marketplace.json)

Claude Code Integration

Claude Code discovers external skills through a manifest-based plugin system. The repository provides a pre-configured manifest that declares the skill's capabilities to Claude.

Step 1: Verify Plugin Manifest Structure

Claude Code reads .claude-plugin/marketplace.json at startup. The manifest maps to the same HTTP endpoint pattern used by the OpenAI integration:

// .claude-plugin/marketplace.json (structure)
{
  "name": "gpt-image-2-style-library",
  "description": "Prompt templates for GPT-Image-2 image generation",
  "type": "http",
  "url": "https://your-deployment.com/gpt-image-2-style-library/generate",
  "input_schema": {
    "type": "object",
    "properties": {
      "templateId": { "type": "string" },
      "customizations": { "type": "object" }
    },
    "required": ["templateId"]
  },
  "output_schema": {
    "type": "object",
    "properties": {
      "prompt": { "type": "string" }
    }
  }
}

Step 2: Deploy the Skill Endpoint

The skill exposes an HTTP server through bin/install.mjs. Deploy this to a URL accessible by Claude Code:

cd agents/skills/gpt-image-2-style-library
npm install
npm run start  # Starts HTTP server on configured port

Step 3: Invoke from Claude Code

Once registered, reference the skill in any Claude Code conversation:

{
  "name": "gpt-image-2-style-library",
  "input": {
    "templateId": "ui-screenshot-system",
    "customizations": {
      "platform": "iOS",
      "aspectRatio": "16:9",
      "title": "My New App"
    }
  }
}

Claude Code routes the request to your deployed endpoint and returns the generated prompt string for use in subsequent image generation calls.

Codex Integration

Codex integrates with gpt-image-2-style-library as a standard npm dependency, enabling direct function calls within your codebase.

Installation

npm install @gpt-image-2/style-library

The package resolves to the local path defined in agents/skills/gpt-image-2-style-library/package.json.

Programmatic API Usage

Import and call generatePrompt from bin/generate.mjs (the package's main entry point):

import { generatePrompt } from '@gpt-image-2/style-library';

async function createImagePrompt() {
  const prompt = await generatePrompt({
    templateId: 'character-design-sheet',
    customizations: {
      character: 'Cyberpunk Hacker',
      poses: 4,
      outfit: 'Neon Jacket'
    }
  });

  // Use with Codex or any OpenAI image model
  const response = await openai.images.generate({
    model: 'gpt-image-2',
    prompt,
    n: 1,
    size: '1024x1024'
  });

  return response.data[0].url;
}

createImagePrompt();

Direct Data Access

For advanced use cases, access the raw style library data:

import styleLibrary from '@gpt-image-2/style-library/data/style-library.json';

// List available templates
const templates = Object.keys(styleLibrary.templates);

// Access specific style metadata
const style = styleLibrary.categories.photography.styles['cinematic-lighting'];

Shared Data Model

Both integrations consume data/style-library.json as the canonical source. This file defines:

Categories: Top-level groupings (photography, illustration, UI/UX)
Styles: Visual aesthetics within each category
Scenes: Compositional contexts
Templates: Pre-built prompt structures with substitution variables

The generatePrompt function in bin/generate.mjs merges template strings with customizations objects to produce final prompts. This logic is provider-agnostic—Claude Code calls it via HTTP, Codex calls it as a library function.

Configuration Comparison

Aspect	Claude Code	Codex
Discovery mechanism	`.claude-plugin/marketplace.json` manifest	`package.json` dependency
Invocation style	HTTP POST to deployed endpoint	In-process function call
Entry point	`bin/install.mjs` (server mode)	`bin/generate.mjs` (library mode)
Authentication	Bearer token in manifest headers	Environment variables in host app
Response format	JSON with `prompt` field	JavaScript string return

Key Files Reference

File	Purpose
[`data/style-library.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/data/style-library.json)	Master data catalog—consumed by both integrations
[`agents/skills/gpt-image-2-style-library/package.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/agents/skills/gpt-image-2-style-library/package.json)	NPM package definition with `main` pointing to `bin/generate.mjs`
[`agents/skills/gpt-image-2-style-library/SKILL.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/agents/skills/gpt-image-2-style-library/SKILL.md)	Human-readable skill documentation for marketplaces
[`agents/skills/gpt-image-2-style-library/agents/openai.yaml`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/agents/skills/gpt-image-2-style-library/agents/openai.yaml)	Reference manifest for converting to Claude format
[`.claude-plugin/marketplace.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/.claude-plugin/marketplace.json)	Claude Code plugin registration
`agents/skills/gpt-image-2-style-library/bin/generate.mjs`	Core prompt generation logic
`agents/skills/gpt-image-2-style-library/bin/install.mjs`	HTTP server setup for Claude Code deployment

Summary

gpt-image-2-style-library provides a unified prompt template system for GPT-Image-2 through [data/style-library.json](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/data/style-library.json).
Claude Code integrates via the [.claude-plugin/marketplace.json](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/.claude-plugin/marketplace.json) manifest, which declares an HTTP endpoint served by bin/install.mjs.
Codex integrates by installing @gpt-image-2/style-library and importing generatePrompt from bin/generate.mjs for direct programmatic access.
Both approaches use identical template substitution logic and consume the same style-library.json source, ensuring consistent prompt output across platforms.

Frequently Asked Questions

What is the minimum required configuration for Claude Code integration?

Claude Code requires only a valid .claude-plugin/marketplace.json file with name, type: "http", url, and input_schema/output_schema fields. The skill HTTP endpoint must accept templateId and optional customizations parameters, returning a JSON object with a prompt string.

Can I use the same deployment for both Claude Code and Codex?

Yes. Deploy bin/install.mjs as a persistent HTTP service for Claude Code, while also publishing the same package to a private npm registry for Codex. Both consume data/style-library.json identically, so updates to the data automatically propagate to both integrations.

How do I add custom templates to the library?

Add new template definitions to data/style-library.json following the existing schema structure, then reference the new templateId in either your Claude Code manifest calls or Codex generatePrompt() invocations. No code changes are required in the integration layers.

Have a question about this repo?

These articles cover the highlights, but your codebase questions are specific. Give your agent direct access to the source. Share this with your agent to get started:

Share the following with your agent to get started:

curl -s "https://instagit.com/install.md"

Add to your MCP client configuration:

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

Ask your agent:

"Use Instagit MCP to understand how freestylefly/awesome-gpt-image-2 works."

Works with

Claude Codex Cursor VS Code OpenClaw Any MCP Client

Maintain an open-source project? Get it listed too →