Codex PPT Skill is a PowerPoint generation tool designed as an agent skill for Codex and other skill-compatible agents like Claude Code and OpenClaw.
The tool converts articles, reports, papers, and course notes into full-page image-based presentations. It works by first planning an outline and visual style, then generating slide images using image generation models, and finally assembling them into a .pptx file using a local script. The approach supports multiple image generation backends including GPT-Image-2.5-Flare, third-party image APIs, and OpenAI-compatible endpoints, making it flexible across different agent environments.
Developers should choose this tool if they need visually rich presentations generated from text content, particularly when strong visual expression matters more than direct editability of individual page elements. The generated slides are image-based rather than editable text, which suits batch generation workflows. For projects requiring further editing of generated presentations, the tool's documentation mentions a companion skill for converting image-based slides into editable PowerPoint format. The skill is designed to work both within Codex's native image generation and as a fallback through external APIs, accommodating different deployment scenarios.
The maintainers typically respond to new issues and pull requests within hours.