image-prompt-reverse

Upload a reference image and get its composition, colour, lighting and style back as Chinese and English image prompts.

Ecosystem: Agent Skills
Skill category: Design
Creator / organization: LunarXuan

What it is

image-prompt-reverse is a Skill by GitHub user LunarXuan. The author says it generates high-fidelity prompts, ready for AI image tools, from a target image the user uploads. It breaks down the reference image's composition, colour, medium and lighting, writes prompts in Chinese and English, and adds a negative prompt. The author built it for Codex: copy it into Codex's Skills folder and call it. What sets it apart is that any text inside the image is treated as something to analyse, never as an instruction to follow.

Who it's for:AI art fans who see an image they love and want to create something in a similar style.

Before you start: you need an AI agent that can read images; the author targets Codex. Images are sent to whichever AI service you use, so it doesn't run offline. If the reference shows real people, brands or someone else's work, reverse-engineering a prompt doesn't give you rights to the original. It is licensed under GPL-3.0.

Based on the project's public documentation. Not tested by AI Inspiration World.