Gpt Image 2

{
  "name": "gpt-image-2",
  "version": "1.0.3",
  "category": "Image Generation / Prompt Engineering",
  "description": "Focused image generation and editing skill for GPT Image 2 and OpenAI-compatible image APIs. Supports three runtime modes — Garden local, host-native delegation, and advisor-only — with 18 visual categories and 80+ structured prompt templates.",
  "homepage": "https://github.com/ConardLi/garden-skills/tree/main/skills/gpt-image-2",
  "compat": [
    "claude-code",
    "claude-ai",
    "cursor",
    "codex-cli",
    "gemini-cli",
    "opencode"
  ]
}


# GPT Image 2 Skill

**A focused image-generation / editing skill for GPT Image 2, with a single SKILL definition that adapts to three runtime modes — local generation, host-native delegation, and pure prompt advisor.**

[中文文档](./README.zh-CN.md) · [Back to collection root](../../README.md)

![GPT Image 2 Skill](https://cdn.jsdelivr.net/gh/ConardLi/assets@main/imgs/gpt-image-2-skill.webp)

---

## What it does

This skill is a structured prompt-engineering and image-generation pack built around the GPT Image 2 model (and OpenAI-compatible image endpoints). It only does two image tasks — `POST /images/generations` and `POST /images/edits` — but it does them in three different runtime environments without changing user-facing behavior.

It bundles:

- A **mode-aware workflow** so the same skill works whether the agent itself owns the image API key, the host has its own image tool, or there is no image tool at all.
- A **structured template library** of 18 categories and 79 prompt templates covering posters, UI mockups, product visuals, infographics, academic figures, technical diagrams, comics, avatars, and editing workflows.
- **Reproducible prompt + image archival** under `garden-gpt-image-2/prompt/` and `garden-gpt-image-2/image/` with task-slug + timestamp naming.

---

## The three runtime modes

The very first thing this skill does on any task is run a tiny detection script:

```bash
node skills/gpt-image-2/scripts/check-mode.js
# or for structured output:
node skills/gpt-image-2/scripts/check-mode.js --json
```

The output picks one of three modes:

| Mode | Trigger | Behavior |
|---|---|---|
| **A — Garden local** | `ENABLE_GARDEN_IMAGEGEN` truthy **AND** `OPENAI_API_KEY` present | End-to-end: pick template → render prompt → call `generate.js` / `edit.js` → image lands on disk |
| **B — Host-native** | Garden disabled, but the host agent already has an image tool (`image_generation`, `dalle`, `nano_banana`, image MCP, etc.) | Render the prompt, then **delegate** image generation to the host's own tool |
| **C — Advisor** | Garden disabled, host has no image tool | Skill degrades into a high-quality prompt writer — saves the rendered prompt to `garden-gpt-image-2/prompt/` and instructs the user to paste it into ChatGPT / Midjourney / DALL·E / Sora / Nano Banana / their own gateway |

In all three modes, prompt files are saved (mode A & C must save, mode B is recommended for reuse). Only mode A produces an image file; mode B leaves that to the host, mode C cannot.

---

## Quick start

### 0. Detect the mode (always step 0)

```bash
node skills/gpt-image-2/scripts/check-mode.js
```

The commands below (1–4) only apply in **Mode A**.

### 1. Text-to-image

```bash
node skills/gpt-image-2/scripts/generate.js \
  --prompt "A cute baby sea otter" \
  --size 1024x1024 \
  --quality high
```

### 2. Generate from a saved prompt file

```bash
node skills/gpt-image-2/scripts/generate.js \
  --promptfile garden-gpt-image-2/prompt/poster-20260424-153045.md
```

### 3. Edit an existing image

```bash
node skills/gpt-image-2/scripts/edit.js \
  --image assets/source.png \
  --prompt "Replace the background with a clean studio scene"
```

### 4. Mask-based local edit

```bash
node skills/gpt-image-2/scripts/edit.js \
  --image assets/source.png \
  --mask  assets/mask.png \
  --prompt "Replace only the masked area with a glass vase"
```

For Mode B / C there is no CLI