google-gemini/gemini-skills — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a lightweight open source skills repo for people using coding agents with the Gemini API.

The goal is to give agents extra up to date guidance so they write correct Gemini API code instead of guessing from old knowledge. I want a clear README that explains what skills are, why they help, how to install them with the Vercel Skills CLI or Context7, and how to connect the Gemini API docs MCP server for current docs.

Create separate skills for general Gemini API app development, real time Gemini Live API apps with audio, video, text streaming and WebSockets, and Gemini Interactions API apps with chat, streaming, function calling, structured output, image generation, and research style agents.

Keep it simple and documentation focused. This doesn’t need to be a web app. It should feel like a practical library that someone can install into Cursor or another coding assistant so the assistant follows current Gemini best practices. Look up current Gemini docs online if you need to.

Want more depth? Deep Reverse