What problem does it solve? It lets you generate images through a self-hosted CliRelay OpenAI-compatible image endpoint (gpt-image-2) when the default system image generation is unavailable or you explicitly want to route requests through your own configured service. ## Core Features & Use Cases - Custom Endpoint Image Generation: Calls a configured CliRelay /v1/images/generations endpoint with gpt-image-2, supporting size, quality, and model overrides. - Credential and Transport Safety: Reads CLIRELAY_API_KEY only from the environment, never logs or persists it, redacts it from error output, and requires an explicit --allow-insecure-http acknowledgement before using the plain HTTP endpoint. - Safe File Output: Refuses to overwrite existing files, validates the returned PNG/JPEG/GIF/WebP format against the file extension, and writes images atomically with 0600 permissions. - Use Case: You have a CliRelay relay with image quota and want a blog illustration saved to a specific path; run a dry-run to validate the prompt, then generate one 1024x1024 high-quality image. ## Quick Start Use the clirelay-imagegen skill to generate a 1024x1024 image from my prompt and save it to a new local PNG file.