# How to Integrate GPT-Image2 Agent Skill with Claude Code: Complete Setup Guide

> Integrate GPT-Image2 Agent Skill with Claude Code for style-aware image prompt generation. Follow this complete setup guide and enable powerful AI art tools in your chat.

- Repository: [苍何/awesome-gpt-image-2](https://github.com/freestylefly/awesome-gpt-image-2)
- Tags: how-to-guide
- Published: 2026-08-26

---

**Install the GPT-Image2 Agent Skill via Claude Code's plugin marketplace commands or npm to enable style-aware image prompt generation directly in your chat interface.**

The **GPT-Image2 style-library** skill allows Claude Code to generate production-ready image prompts by tapping into the shared [`style-library.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/style-library.json) dataset from the freestylefly/awesome-gpt-image-2 repository. Once integrated, you can summon visual styles, prompt templates, and scene tags without leaving your development environment.

## What is the GPT-Image2 Style-Library Skill?

The skill functions as a bridge between Claude Code and the curated visual style database used by the GPT-Image2 web gallery. According to the source code in [`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), the skill exposes a workflow that detects your request language, identifies the target output type (infographic, poster, portrait), and matches the most appropriate template from [`data/style-library.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/data/style-library.json).

Key capabilities include:

- **Automatic style matching** against 200+ curated templates
- **Bilingual support** for Chinese and English prompt generation
- **Local execution** via the `bin/install.mjs` entry point

## Installation Methods

You can integrate the skill using either Claude Code's native plugin marketplace or manual npm installation for development workflows.

### Method 1: Installing via Claude Code Plugin Marketplace

Claude Code discovers skills by scanning the `~/.claude/skills` directory for folders containing a [`SKILL.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/SKILL.md) file. To install via the marketplace:

1. **Register the repository:**

   ```text
   /plugin marketplace add freestylefly/awesome-gpt-image-2
   ```

2. **Install the specific skill:**

   ```text
   /plugin install gpt-image-2-style-library@awesome-gpt-image-2
   ```

3. **Reload the session:**

   Restart Claude Code or execute `:reload` to scan `~/.claude/skills` and register the new skill metadata.

4. **Verify installation:**

   ```text
   /skill list
   ```

   You should see `gpt-image-2-style-library` in the available commands.

### Method 2: Manual npm Installation for Developers

For local development or testing unpublished versions, use the npm package directly:

```bash
npm install -g gpt-image-2-style-library
gpt-image-2-style-library install all

```

The `install all` command executes `bin/install.mjs`, which copies the skill into all supported local agent directories: `~/.claude/skills`, `~/.codex/skills`, and `~/.agents/skills`. After installation, restart Claude Code to trigger the skill discovery mechanism.

## How Claude Code Consumes the Skill

Understanding the internal flow helps debug integration issues and customize templates.

1. **Discovery phase:** At startup, Claude Code recursively scans `~/.claude/skills` for directories containing a [`SKILL.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/SKILL.md) file.

2. **Metadata parsing:** The YAML front-matter in [`SKILL.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/SKILL.md) (fields `name` and `description`) registers the skill identifier `gpt-image-2-style-library`.

3. **Command routing:** When your message begins with the skill name, Claude Code forwards the request to the compiled JavaScript module specified in the skill's `bin/install.mjs` entry point.

4. **Data access:** The runtime imports the bundled [`data/style-library.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/data/style-library.json) to perform category and template lookups against the latest gallery cases.

5. **Response formatting:** The skill constructs the final prompt following the structured sections defined in [`SKILL.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/SKILL.md) (Subject, Composition, Style, Negative Constraints) and returns plain text to the chat interface.

## Using the Skill in Practice

Once integrated, invoke the skill naturally in conversation or via the explicit skill command.

### Example 1: Generating an Infographic Prompt (Chinese)

Request:

```text
/skill gpt-image-2-style-library
用 gpt-image-2-style-library 技能生成关于 "Codex" 的信息图

```

Claude Code processes the Chinese input, detects the "infographic" intent, and queries [`style-library.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/style-library.json) for matching templates. The response follows the exact output format specified in [`SKILL.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/SKILL.md):

```

Prompt:
Codex 信息图，包含以下模块：
- 标题："Codex 代码生成平台概览"
- 子标题：简要说明功能
- 结构化箭头展示工作流：输入 → 处理 → 输出
- 颜色方案：淡蓝 + 灰白
- 视觉风格：现代平面设计，圆角卡片
- 纵横比：16:9
- 负面约束：避免过度装饰，保持简洁

参考案例：#123、#210（相似信息图）

```

### Example 2: Creating a Product Poster (English)

Request:

```text
/skill gpt-image-2-style-library
Create a product poster prompt for a new AI-powered coffee mug.

```

The skill identifies the "product poster" category from [`data/style-library.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/data/style-library.json) and returns:

```

Prompt:
AI Coffee Mug – product poster
- Subject: sleek matte-black mug with embedded AI display
- Layout: central product, top tagline, bottom specs
- Style: minimalistic, high-contrast, neon teal accents
- Text: "Smart Brew. Smarter You."
- Aspect ratio: 4:5
- Constraints: no background clutter, focus on surface texture

Template: poster-standard-v2 (see style-library)

```

## Key Source Files and Architecture

The integration relies on these specific files from the freestylefly/awesome-gpt-image-2 repository:

- **[`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)** – Defines the skill metadata, workflow steps, and output defaults that Claude Code parses at startup.

- **`agents/skills/gpt-image-2-style-library/bin/install.mjs`** – CLI entry point that copies skill files to local agent directories (`~/.claude/skills`, etc.) and handles the `install all` command.

- **[`data/style-library.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/data/style-library.json)** – The consolidated database of templates, category tags, and example cases shared between the web gallery and the agent skill.

- **[`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 manifest specifying the global binary name and version metadata.

When the gallery updates, regenerate the skill using `npm run generate:style-skill` followed by `npm run install:skill` to propagate new templates to Claude Code.

## Summary

- The GPT-Image2 Agent Skill connects Claude Code to the [`style-library.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/style-library.json) template database via the `gpt-image-2-style-library` npm package.
- Install via `/plugin install` commands or `npm install -g gpt-image-2-style-library` followed by `install all`.
- Claude Code loads skills from `~/.claude/skills` by parsing [`SKILL.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/SKILL.md) metadata at startup.
- Invoke the skill with `/skill gpt-image-2-style-library` followed by your image generation request in English or Chinese.
- Update the skill by running `npm run generate:style-skill` and reinstalling to refresh templates from the gallery.

## Frequently Asked Questions

### Where does Claude Code store the GPT-Image2 skill files?

Claude Code stores skills in the `~/.claude/skills` directory. The `bin/install.mjs` script copies the [`SKILL.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/SKILL.md) and supporting assets into a subfolder named `gpt-image-2-style-library` within that path. Claude Code scans this location at startup to build the available command list.

### Can I use this skill with other agents besides Claude Code?

Yes. The installation script targets multiple agent directories: `~/.claude/skills`, `~/.codex/skills`, and `~/.agents/skills`. Running `gpt-image-2-style-library install all` deploys the skill to all compatible agents simultaneously, provided they follow the same [`SKILL.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/SKILL.md) discovery convention.

### How do I update the skill when new templates are added to the gallery?

Run `npm run generate:style-skill` in the repository root to regenerate [`data/style-library.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/data/style-library.json) from the latest gallery cases, then execute `npm run install:skill` to copy the updated files to `~/.claude/skills`. Restart Claude Code to load the new templates.

### What information does the skill use to build prompts?

The skill parses [`data/style-library.json`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/data/style-library.json) to extract subject classifications, composition guidelines, color schemes, and negative constraints. It matches your request against these fields using the workflow defined in [`SKILL.md`](https://github.com/freestylefly/awesome-gpt-image-2/blob/main/SKILL.md) to construct a complete image-generation prompt with proper aspect ratios and style tags.