00 — Introduction

In this chapter:

  • What an Agent Skill is and why it matters
  • What Dr JSkill generates, and how it differs from tools you may know (JHipster, Spring Initializr)
  • The mindset shift when you code with an AI agent instead of by yourself

No commands to run here — just ideas. Chapter 1 is where the hands-on starts.

1. What is an Agent Skill?

An AI coding agent (GitHub Copilot CLI, Claude Code, Cursor, etc.) is a program that can read your files, run commands, and edit your code — all driven by natural-language instructions. Give it a task, it figures out the steps.

Agents are powerful but generic. They know "a bit of everything" and, left alone, tend to produce code that is:

  • Plausible but idiosyncratic — every run looks slightly different
  • Mixed in quality — great for one file, surprising for another
  • Disconnected from your team's or your community's conventions

An Agent Skill is a small, portable bundle of instructions — Markdown files, scripts, and references — that teaches an agent how you want a specific kind of task done. It's a specification the agent reads before it starts working.

Think of it as a "senior engineer's checklist" that travels with the agent.

2. What Dr JSkill is

Dr JSkill is an Agent Skill that teaches any compatible AI coding agent how to generate Spring Boot applications following Julien Dubois' best practices.

When you ask the agent to "create a new Spring Boot app", Dr JSkill tells it:

  • Use Spring Boot 4.x with Java 25
  • Use Maven (not Gradle), Hibernate ddl-auto (not Flyway/Liquibase)
  • Use PostgreSQL in Docker for development
  • Wire a front-end (Vue, React, Angular, or vanilla JS) through the Maven Frontend Plugin so ./mvnw package builds the whole thing
  • Ship sensible defaults: .gitignore, .editorconfig, .env.sample, a Dockerfile, a compose.yaml, CI config, etc.
  • Avoid a specific list of things the author considers traps: Lombok, Gradle, OpenAPI/springdoc scaffolding, buildpacks, Jib

The result: different agents, different runs, and different prompts all produce applications that look like each other and look like what a senior Spring Boot developer would write by hand.

3. How this differs from Spring Initializr

Spring Initializr (start.spring.io) gives you a skeleton project — pom.xml, an empty Application.java, and nothing else. You still have to write every controller, entity, and configuration yourself.

Dr JSkill uses Spring Initializr under the hood to bootstrap the project, then asks the agent to fill in the rest: domain entities, REST controllers, a working front-end, database configuration, tests, Docker, and the "dotfiles" that make a project feel professional.

4. The mindset shift

If you're new to AI coding agents, expect three things to feel different:

  1. You describe outcomes, not steps. Instead of writing the code for a user filter, you write: "Add a dropdown at the top of the todo list to filter by user." The agent figures out which files to edit.

  2. You review diffs, not code you wrote. Your main job during the workshop is to read what the agent changed (via git diff) and decide whether it matches what you asked for. Dr JSkill tilts the odds in your favor, but the agent is still not infallible.

  3. git is your undo button. Nothing the agent does is sacred. If a change goes sideways, git restore . erases it and you try a different prompt. Make frequent commits — you'll want those checkpoints.

5. What you'll learn

By the end of this workshop you'll be able to:

  • Bootstrap a real Spring Boot application with a few sentences of English
  • Read a generated project and understand why each piece is there
  • Extend an existing project by talking to the agent, safely and iteratively
  • Test, tune, and deploy what you built
  • Apply the same workflow to your own projects — with or without Dr JSkill

6. Tested models

This workshop has been executed end-to-end (chapters 02 → 08) with three AI models, all of them available in Copilot CLI:

  • Claude Haiku 4.5
  • Claude Opus 4.7
  • GPT-5.5

All three produced a running application, but the quality of the generated code varied. Dr JSkill works best with GPT-5.5, which is the recommended model for the workshop. Chapter 2 gives concrete advice on selecting it when you start your session.

Checkpoint

Nothing to run in this chapter. If you're attending a facilitated workshop, this is a good moment to pause for questions.

Next → Chapter 1 — Setup