wwe-dog/ppt-image2-editable-rebuild

This is a Codex skill, meaning a set of instructions that an AI coding assistant can follow as a slash command. Its purpose is to help turn AI-generated slide reference images into editable PowerPoint files. The workflow has two stages: first generate a visual mockup of each slide using an image generation tool, then use this skill to rebuild that mockup as a real PPTX file with editable text boxes, shapes, and layout elements. The core idea is that pasting a generated image directly into a slide produces a picture with no editable content. This skill instead treats the reference image as a blueprint and splits each slide into two kinds of content. Complex visual elements such as illustrations, icons, maps, and decorative backgrounds are kept as cropped screenshots taken from the reference image. Structural elements such as titles, body text, number cards, arrow callouts, and conclusion bars are rebuilt as native PowerPoint shapes that can be edited or restyled. The readme also documents guidance for generating better reference images: keep backgrounds clean and simple so that when sections are cropped out for visual assets, the text and card regions are easy to isolate. Complex or heavily textured backgrounds make the cropping step harder and leave residual artifacts in the output slide. The skill includes a Python script for cropping asset regions from reference images and an example set of 29 slide reference images with matching prompt documents. The repository and readme are written in Chinese. The tool is designed for use within the Codex AI coding assistant environment and is not a standalone desktop application.