Prompt Wizard

AI-powered prompt generator for ChatGPT Image 2. Describe your vision, get a professional-grade English prompt.

Demo

Click the GIF to watch the full MP4 demo with audio.

Also check the Landing Page for a full feature overview.

---

Quick Start

/wizard a cyberpunk cat sitting on a neon-lit balcony overlooking a rainy Tokyo street

Prompt Wizard asks 2-5 clarifying questions (text/copy, market, style preferences, key details), then outputs a polished GPT-Image-2 prompt in English — the optimal language for image generation quality.

Commands

Command	Description	Example
/wizard [idea]	Conversational generation with style anchoring	/wizard luxury watch ad
/wizard pro	Structured 8-step guided mode (/back to revise, /skip to accept default)	/wizard pro
/wizard multi [idea]	Generate multiple prompts in one session (batch, variations, series)	/wizard multi 3 variations of a luxury watch ad
/wizard examples	Showcase curated highlight cases with local image previews	/wizard examples
/wizard templates [cat]	Browse 7 categories of community-verified templates	/wizard templates portrait
/wizard lang [code]	Switch conversation language (prompts always English)	/wizard lang ja
/wizard update-library	Fetch latest prompt library from upstream	/wizard update-library
/wizard version	Show skill and library version info	/wizard version

Refine after generation: darker, brighter, more minimalist, closer crop, warmer tones, etc.

Style anchoring in /wizard: The skill searches the case library for matching references and anchors your prompt to a specific case's lighting, composition, or palette — this step is part of the default workflow and must be confirmed before generation.

Multi-prompt mode (/wizard multi): For 4+ prompts or narrative series, generates the first prompt for your review before producing the rest — ensuring the style direction is locked in.

Market dimension: Both modes ask about target market/audience (Chinese, Japanese, Korean, Western, global) to align the aesthetic direction.

Supported conversation languages: zh, en, ja, ko, es, fr, de, ru, pt

Install

Claude Code

Git (recommended — full install with 175+ case images, ~90MB):

git clone https://github.com/EricArcha/prompt-wizard.git ~/.claude/skills/prompt-wizard

npx clawhub (lightweight, ~8MB):

npx clawhub@latest install prompt-wizard

OpenClaw

openclaw skills install (recommended — lightweight, ~8MB):

openclaw skills install prompt-wizard

After install, run /wizard update-library to download the full case library with 175+ reference images.

Git (full install with 175+ case images, ~90MB):

git clone https://github.com/EricArcha/prompt-wizard.git ~/.openclaw/skills/prompt-wizard

Update the prompt library

The case library is bundled with the skill. To fetch the latest community prompts:

To update the skill itself (new features, bug fixes):

cd ~/.claude/skills/prompt-wizard && git pull

To update the case library (latest community prompts):

/wizard update-library

This fetches the latest version of awesome-gpt-image-2-prompts and updates the version in config.json.

How It Works

Prompt Wizard is built on analysis of 175+ community-verified GPT-Image-2 prompts. It encodes the patterns that consistently produce high-quality results:

1. Market-aware direction — aligns aesthetic to target audience (Chinese/Japanese/Korean/Western/global)
2. Style anchoring — matches your vision to proven case patterns before generation
3. 6-dimension coverage — Subject → Environment → Lighting → Composition → Style → Technical
4. English output — all prompts generated in English for optimal ChatGPT Image 2 performance
5. Case-grounded generation — every prompt is anchored to at least one community-verified case from the library; Related Cases section is mandatory in output

File Structure

prompt-wizard/
├── SKILL.md                      # Core skill logic
├── README.md / README_zh.md      # Docs
├── demo/                         # Demo animation + landing page
│   ├── demo-animation-bgm.mp4    # Product demo video (with audio)
│   ├── demo-animation.html       # Source HTML animation
│   └── landing.html              # Feature landing page
├── references/
│   ├── prompt-structure.md       # 6-dimension reference
│   └── category-guides.md        # Category-specific tips
├── scripts/
│   └── update-prompts.sh         # Library update script
├── data/
│   └── awesome-gpt-image-2-prompts/  # Bundled case library (~88MB, 500+ files)
└── config.json                   # Version + preferences

Credits

• Case library: awesome-gpt-image-2-prompts (CC BY 4.0) by EvoLinkAI and community
• Skill design: built for Claude Code and OpenClaw

Roadmap

• Conversational + pro mode generation
• Style anchoring with case library
• Bundled case library with local images
• Bilingual documentation (EN/ZH)
• ClawHub registry publication
• Multi-image prompt support (sequences, grids, storyboards)
• User prompt history / favorite collection

License

Personal & educational use is free. Commercial use requires authorization. See LICENSE for full terms.

Note: data/awesome-gpt-image-2-prompts/ is CC BY 4.0 (see LICENSE).

Packages installed from ClawHub use the MIT-0 license (platform requirement). This repository uses a custom license.

---

Prompt Wizard · Built by EricArcha
_{Available on Claude Code & OpenClaw}