The Agentic Crew: Hands-On Guide

What's inside

Three hands-on exercises

The Hands-On Guide is the practical companion to the main engineering book. Where the main book explains the ideas, this one puts them into practice.

You don't need to be a programmer. You need to be willing to type commands into a terminal and see what happens. That's how you learn this — not by reading, but by doing.

Setting Up Your Workstation

Terminal, package manager, Git, GitHub CLI, and an AI coding assistant — from scratch, on any OS.

Your First Pull Request

Fork a real repository, write a review, commit and push, and submit a pull request that goes live on GitHub.

Your First AI-Generated Project

Fork the web-presenter project and use an AI agent with a Hugging Face API key to generate custom illustrations, spoken narration, and animated backgrounds — all from prompts.

→

More chapters coming

Pair programming with an agent, building from scratch, adding guardrails, multi-agent workflows, and a capstone project.

Contents

3 exercises — more coming

Setting Up Your Workstation Terminal, package manager, Git, GitHub CLI, and an AI coding assistant — from scratch. Each edition is tailored to your OS. Ends with a working environment and a verified connection to GitHub.

Your First Pull Request Fork a real repository, read a chapter, write a review, commit and push, and submit a pull request that goes live on GitHub. The full open-source workflow in one exercise.

Your First AI-Generated Project Fork the web-presenter project and use an AI agent with a Hugging Face API key to generate custom illustrations, spoken narration, and animated backgrounds — all by describing what you want in plain English.

Pair Programming with an Agent Coming soon — work on a real bug in a real codebase alongside an AI agent, using the agent as a thinking partner rather than just a code generator.

Prompts That Work Coming soon — a practical workshop in writing prompts that get consistent, useful results. Anti-patterns, decomposition, and the prompt-as-spec technique.

Building from Scratch Coming soon — start with a blank folder and a one-sentence brief, and build a complete working project using an AI agent from the first line to deployment.

Adding Guardrails Coming soon — take an existing project and add approval gates, permission scoping, and environment-specific rules so the agent can work safely without supervision.

Multi-Agent Workflows Coming soon — run two agents in parallel on separate branches, then review and merge their work. Branch-per-agent isolation in practice.

CI/CD with Agents Coming soon — wire an agent into a GitHub Actions pipeline for automated PR pre-screening and overnight test runs.

Capstone Project Coming soon — a full project from brief to deployed product, using everything covered in the previous exercises.

Exercise highlights

What you'll actually do

⌨️

Run real commands

Every step uses actual terminal commands — git, gh, python3. No web interfaces, no drag-and-drop.

🌐

Work with live GitHub repos

You'll fork schlunsen/theagenticcrew and schlunsen/web-presenter — real projects with real histories.

🖼️

Generate AI illustrations

Use FLUX.1-schnell via the Hugging Face API to generate ten custom images for your presentation — from plain English descriptions.

🔊

Generate AI narration

Turn your slide text into spoken audio using text-to-speech models on Hugging Face — no recording required.

✨

Customise animated backgrounds

Ask an AI agent to change the Three.js network animation to match your presentation's colour palette — in seconds.

📤

Ship it to GitHub

Every exercise ends with a commit and push. Your work lives on GitHub — shareable, versioned, real.

Learn by doing, not by reading