This collection provides prompts and guidance to build AI agents with Claude Code, focusing on automating email triage. Departing from 'black box' solutions like OpenClaw, this collection emphasizes transparency and control by building solutions with Claude Code and Bash. You’ll get hands-on experience with all four key zones of an AI agent.

After working through the prompts, you'll be able to construct custom email triage agents that can be triggered via Telegram, access your Gmail account, summarize emails based on different criteria, and deliver briefings. You’ll be able to do that without reliance on opaque abstraction layers.

This collection is designed for developers, automation enthusiasts, and anyone seeking to build programmable, transparent AI agents with full control. It's particularly useful for those who are already familiar with basic scripting and want to extend the power of LLMs, using Claude Code.

You will learn how to design agents using the 4 zone architecture. You will also learn how to create modular agents that can be easily modified and extended.