Codex PPT Skill converts articles and notes into image-based PPT slides using gpt-image-2, supports multiple agents, and assembles slides into PPTX with speech notes. Latest releases add style references and 2K/16:9 defaults, with 4K option guidance.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Codex PPT Skill is a PPT-creation tool that converts articles, reports, outlines, or notes into image-based slide decks. It supports Codex, Claude Code, OpenClaw, Hermes Agent, and other SKILL.md-compatible environments. It plans outlines, generates slide images, and assembles them into a PPTX, optionally recording a speech.md per slide.
How it works
The workflow includes reading content, outlining, confirming page count and points, proposing visual styles, generating slide images per page, assembling into PPTX, and producing speech notes. It can guide users to save preferred styles to a local style library and supports multi-agent concurrent generation with per-page checks.
Getting started
One-line install (recommended)
请帮我安装这个 codex-ppt skill，链接是：https://github.com/ningzimu/codex-ppt-skill
Manual installation to Codex
npx -y skills@latest add ningzimu/codex-ppt-skill \
--skill codex-ppt \
--agent codex \
--global
Other installation pathways include OpenClaw and Claude Code/Hermes, and there are instructions for linking the local repository for development.
Updates
请帮我更新 codex-ppt skill 到最新版本，仓库是：https://github.com/ningzimu/codex-ppt-skill
API keys and style libraries are stored outside the skill directory to survive updates.
Recent releases
- v0.5.5 (2026-07-13): Features include storing user custom styles outside the skill install in a references directory with automatic discovery.
- v0.5.4 (2026-07-12): Added a Party-and-Government Red style reference and a Teaching Courseware style reference.
- v0.5.3 (2026-07-09): Simplified speaker notes to include only the spoken talk track per slide.
- v0.5.2 (2026-06-20): Reverted to v0.5.0 release state.
- v0.5.1 (2026-06-18): Updated workflow defaults for 2K 16:9 image output, direct sample prompts, mandatory subagent dispatch/refill behavior, and edit-based single-slide revisions.
Documentation: GitHub Pages site linked in README.
Traction
4497 stars, 223 forks, 2 open issues as factual metrics provided.
Output structure
Each PPT creates a project directory with origin_image/, outline.md, speech.md, and a .pptx. Slides are saved as slide_01.png, slide_02.png, etc.
Caveats
License: MIT. Created 2026-04-26; last push 2026-07-30. Open issues: 2. Language: Python. This summary does not add speculative claims beyond the documented facts.