codex-imagen

What problem does it solve?

Codex Imagen solves the friction of generating or editing images by letting agents call the Codex backend directly using OAuth credentials already stored locally, without requiring an OpenAI API key.

Core Features & Use Cases

• Native image_generation calls via Codex Responses: Uses the backend-api/codex Responses image_generation tool for raster image outputs.
• Local OAuth credential discovery + refresh: Automatically locates OpenClaw/Codex auth JSON and refreshes tokens (including recovery from refresh_token_reused).
• Reference-image support for edits/composition: Accepts local images, HTTP(S) URLs, or data URLs and sets input_image.detail for control over fidelity.
• Streaming and multi-output saving for agent workflows: Streams results and writes each completed image to disk, returning paths (or JSON summaries) suitable for OpenClaw and other orchestrators.
• Use Case: An OpenClaw agent needs to produce multiple UI icon variants from prompts and optional reference artwork, then return stable saved file paths to the conversation.

Quick Start

Ask the skill to generate an image and save it to disk by running the helper with your timeout and prompt.