AI Examples — AI skill for Claude Code

Starting with AI coding agents

A starting-point template for engineers who already know systems work and are new to AI coding agents — a guide first: eleven chapters, copy-out templates, and two worked examples. It is **pi-first**, with a Claude Code companion track. The public, package-installable stack is taught as working instructions; the private layer is named as patterns with a smallest-useful build recipe for each, and none of those patterns is installable from here.

**Cloned this to use it?** Start at [`ADOPTING.md`](ADOPTING.md). It sequences the work — which template goes where, in what order, and how to prove the harness actually loaded it — and it ends with a self-contained brief you can hand to an AI agent to do the whole setup for you. You do not need to read the chapters first: they explain *why* the templates have the shape they do, and you can read one when a decision actually bites.

Nobody is updating this as pi, npm, or Claude Code change, and that is deliberate — it is built to be forked, not followed. Commands, package names, and paths in the chapters were true when those chapters were written, so look them up again before you trust them; every chapter that needs a current value gives the command that produces it. A factual error or a broken link is still worth an issue — see [Maintenance posture](#maintenance-posture).

**Failure prevented:** treating a dated guide as a live product, then following a command that no longer matches upstream — and, at the other extreme, a repository that says "here is an example" and leaves you to invent the adoption path yourself.

Route

What you want	Where it lives
Adopt it — what to copy where, in what order, and the AI-agent brief	`ADOPTING.md`
Mental model (tokens, compaction, failure modes)	`docs/01-mental-model.md`
Install and authenticate	`docs/02-install.md`
Configuration, trust, model lookup