Codex GPT Image

What Codex GPT Image does

The skill reads the local ~/.codex/auth.json file and sends requests to the Codex Images backend on chatgpt.com, using either the generations or the edits endpoint. This lets an agent reuse an existing Codex or ChatGPT subscription to generate images. The default model is gpt-image-2.5-flare, with quality settings low, medium, high, xhigh, max and auto, and the script checks that output sizes are valid.

It supports text-to-image, editing from one or more reference images, and mask-based local edits, plus the Images API options background, moderation, output_format, output_compression and mask. Transparent backgrounds work with PNG or WebP output. If no Codex login exists, a device-code flow writes tokens to ~/.codex/auth.json after the user confirms in a browser. The script is a single Python standard library file, and the repository ships one skill under skills/codex-gpt-image with an English README alongside the Chinese one.

Key features

• Image generation through Codex OAuth, with no OPENAI_API_KEY
• Text-to-image, multi-reference editing and mask-based local edits
• Quality settings low, medium, high, xhigh, max and auto, plus size validation
• Support for background, moderation, output_format and output_compression options
• Built-in device-code login when no Codex auth file exists
• Pure Python standard library script

When to use it

• Letting a Hermes Agent produce images using an existing Codex or ChatGPT login
• Editing a reference image or a masked region from an agent session
• Sharing one image skill across OpenClaw, Claude Code, Codex and Hermes

Who it is for: Agent users who already have a Codex or ChatGPT login and want image generation without configuring an OpenAI API key.

How it fits with Hermes Agent

Supports Hermes Agent among other agents: the README names Hermes Agent next to OpenClaw, Claude Code and Codex as agents that can load this SKILL.md.

How to install Codex GPT Image

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

npx -y skills@latest add ningzimu/codex-gpt-image \
 --global

Requirements: A Codex OAuth login (codex login, or the skill's built-in device-code login) and Python; no OPENAI_API_KEY needed

Note: It calls the Codex Images backend on chatgpt.com, which the README says is not an officially recommended OpenAI API integration, may change or stop working, and depends on account and usage permissions.

FAQ