OpenAI's reference shelf for teaching Codex new tricks

What it does

This repo is a collection of official plugin examples for OpenAI’s Codex. Each plugin is a structured bundle—anchored by a plugin.json manifest and optionally fleshed out with skills, MCP configs, hooks, and commands—that teaches the agent how to work with specific tools or workflows.

The interesting bit

The manifest format is the quietly important part. It standardizes how you declare what a plugin does, what surfaces it exposes, and how Codex should invoke it—essentially a contract between human intent and agent behavior. The examples span design tools, productivity suites, and full-stack deployment targets.

Key highlights

• Figma plugin: covers use_figma, Code to Canvas, and design system rules
• Notion plugin: planning, research, meetings, and knowledge capture workflows
• iOS & macOS plugins: SwiftUI/AppKit implementation, debugging, build loops, and packaging
• Web & Expo plugins: deployment, payments, database workflows, plus React Native SDK upgrades and EAS
• MCP-backed bundles: Netlify, Remotion, and Google Slides plugins using the Model Context Protocol

Verdict

Worth browsing if you’re building Codex plugins and want to see how OpenAI structures manifests and skills. Skip it if you’re looking for a general plugin framework—this is reference material, not a reusable engine.