Appearance
Level 1: How to Prompt Coding Agents
The single highest-leverage skill in the age of AI coding is not writing code — it's writing prompts that make agents write the right code.
The journey starts here. Before you can understand how agents work, optimize their context, or scale them across a team, you need one foundational skill: the ability to communicate intent clearly to an AI coding agent. This chapter is your on-ramp.
We structured this chapter around three progressive questions:
What changed? — AI coding agents aren't just faster autocomplete. They fundamentally change how you spend your time as a developer. How Coding Agents Change Development reframes the developer role from writing code to directing agents, and shows you what the new inner loop looks like.
How do I talk to them? — Knowing the landscape isn't enough; you need practical technique. Prompting Principles distills the highest-impact patterns for writing prompts, CLAUDE.md files, and system instructions — giving context before instructions, being specific about constraints, and knowing when to restart a conversation.
How do I make it stick? — Great prompts are wasted if you have to repeat them every session. AGENTS.md introduces the concept of persistent agent memory — a single file that turns "Claude, we use pnpm" from a per-session reminder into institutional knowledge.
What's next: Once you can prompt agents effectively, the natural question is: what's actually happening under the hood? Chapter 2 pulls back the curtain on how coding agents work — their architecture, autonomy levels, and failure modes — so you can move from "it usually works" to "I understand why it works."