GPT Image 2 Skill offers a gallery of prompts, an agentic skill, and a CLI for image generation/editing via OpenAI Image 2. It supports CAC workflows across Codex, Claude Code, and other runtimes, with installations via plugins, npx, or manual methods.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
A GPT Image 2 prompt gallery plus an agentic skill and a CLI for OpenAI image generation and editing. It surfaces reusable prompts for research figures, posters, UI mockups, and related workflows, and provides runnable examples for skill-capable agents.
How it works
The project surfaces a small, curated gallery of image prompts and offers an agentic skill alongside a CLI named gpt-image. It includes multiple installation paths (Claude Code plugin, Codex-npx installer, manual agent-skill install) and a CLI workflow that maps prompts to OpenAI image endpoints (generate and edits). It documents parameters for prompt, image inputs, mask, size, quality, format, and other options, and shows how to invoke the CLI or SDK to generate or edit images using the gpt-image-2 model. It reads OPENAI_API_KEY from environment or .env sources and notes safety considerations for key handling.
Getting started
- Quick start steps are provided for multiple runtimes via details blocks:
- Claude Code: plugin marketplace add and installer commands
- Codex and other runtimes: agent-skill installers via npx or manual git clone
- CLI: commands to run gpt-image after installation, plus a path to install into PATH
- The repository provides usage examples for common tasks:
- Text → image: gpt-image -p "a photorealistic convenience store at 10pm" --size 1k --quality high -f store.png
- Text + reference image → image (edit): gpt-image -p "Make it a winter evening with heavy snowfall" -i chess.png --quality high -f chess-winter.png
- Multi-reference edits and mask-based inpainting are described with respective command patterns
- It includes a complete parameter reference covering flags like -p/--prompt, -f/--file, -i/--image, -m/--mask, --input-fidelity, --size, --quality, -n/--n, --background, --moderation, --format, and --compression
Recent releases
- v0.2.0 — 2026-04-25: Highlights include documentation, gallery, skill-scale, and discoverability updates after the initial public release.
- v0.1.0 — 2026-04-24: Initial Prompt Gallery release with 151 curated prompts / examples and bilingual English/Chinese README coverage.
Traction
- Stars: 4147
- Forks: 356
- Open issues: 4
Behind the repo
- Maintains cross-runtime agent skill integration and multiple installation methods for Codex, Claude Code, OpenClaw, Hermes Agent, and other runtimes via a shared skills folder structure. The repository highlights include a gallery-centric approach and CLI for the gpt-image-2 model.
Caveats
- License: MIT
- Created: 2026-04-22
- Last push: 2026-07-20
- Language: Python
- Open issues: 4
- Documentation notes bilingual English + 中文; various details depend on runtime integration and environment setup