xutowei/image-to-ui

This project provides a single tool called image-to-ui, described as a Codex skill. Codex skills are reusable tasks that an AI coding assistant can invoke by name during a development session. This particular skill takes two inputs: a reference image showing how a game user interface should look, and a directory of sliced sprite files, which are the individual graphical pieces that make up that interface. The tool examines both together and writes a structured file called ui_structure.json that describes the layout as machine-readable data rather than raw pixels. The example in the README shows the tool being invoked by asking the agent to use it with a specific design file and sprite folder, after which it generates the output files automatically. The use case appears to be aimed at game development. Designers typically produce high-fidelity mockups of UI screens and separately maintain folders of cut graphic assets. Converting those into a JSON description of the layout would let a game engine or codebase position each component programmatically, without manually reconstructing the arrangement by hand. Alongside the JSON, the tool also creates a comparison.png file that places the original design next to the tool's interpretation, so you can spot any discrepancies visually before using the output. The README is short and does not describe the internal implementation, the JSON output schema, or how the sprite matching is performed. No installation steps, dependencies, or technical requirements are listed. The project appears to be in an early or minimal-documentation state.