What is slide-image-to-editable-pptx?

This project is a "skill" for AI coding agents such as OpenAI Codex or Claude Code. A skill, in this context, is a set of structured instructions you install so that an AI assistant knows how to perform a specific task. This particular skill teaches the agent to take screenshots of presentation slides and rebuild them as a proper, editable PowerPoint file, where text stays selectable and shapes stay adjustable. The problem it addresses is a common frustration: you have a set of slide images (exported PNGs, screenshots from a PDF, or AI-generated mockups) and you want them back as a real .pptx file, not a deck of flat pictures pasted as slide backgrounds. Asking an AI to "recreate" a slide usually produces something with roughly the right topic but the wrong layout. This skill tries to do an actual reconstruction by analyzing the original image carefully. The skill breaks the work into three phases. In the first phase, the agent inspects each slide image and catalogs every visible element, such as titles, body text, icons, rectangles, and background illustrations, assigning each one to a category. Text and simple shapes are marked to become native PowerPoint objects. Complex visuals, like scientific diagrams or decorative artwork, are marked to be regenerated as clean image files. In the second phase, the agent generates those image files. In the third phase, it assembles everything into a .pptx file with correct layering, then renders previews and checks them against the source. Installation involves copying the skill folder into your agent's skills directory and then driving it with a three-prompt workflow: one prompt per phase. The README explains each prompt in detail and describes what intermediate files get produced between phases (an element inventory, an asset report, and a final validation report). The project is licensed under MIT and built on a JavaScript library called PptxGenJS, which handles the actual file creation. It works with Codex and Claude Code, other AI coding agents that support a similar skill or plugin system may also be compatible.