Agentic Engineering Playbook

From vibe-coded prototype to professional product.

A collection of patterns for building production software with AI coding agents. Each chapter captures hard-won lessons from shipping a real product this way — the kind of knowledge that's easy to develop and easy to forget.

Who it's for

Primary audience: Future-me, starting a new product. Hand these chapters to Claude (or your AI coding agent of choice) at project start and skip the rediscovery phase.

Secondary audience: Builders who have a working prototype — built with Bolt, Lovable, Cursor, or pure vibe coding — and want to turn it into something professional. You don't need "how to prompt an AI." You need "how to run a software operation where AI does most of the coding."

How to use it

With an AI coding agent: Add the relevant chapters to your project context. They're written to work as bootstrap documents — enough context that your agent knows the patterns, pitfalls, and workflows without you re-explaining everything.

As a human reader: Read the foundation chapters in order if you're starting fresh, or jump to the stage chapter that matches where you are.

Foundation chapters (read in order)

These build on each other. Start here.

#ChapterWhat it covers
1Getting StartedGround-clearing decisions and the six foundations contract
2GuardrailsPrevent disasters from humans and LLMs — environment pipeline, database controls, secret hygiene
3Code QualityWhy traditional review fails for AI-generated code, and what to do instead

Stage chapters (pick up when ready)

These are independent. Reach for them when the need arises.

ChapterWhen you need it
Token EconomicsYour AI costs are rising or your context window feels cramped
API Tokens & SecretsYou have your first API key and nowhere safe to put it
Multiplying ThroughputYour single-agent workflow is a bottleneck
Running a Team of AgentsYou need specialized roles, not just more hands
Security FundamentalsYou have users and something worth protecting — layers, attack trees, tooling
Security in the Mythos EraYou depend on open-source packages and probability of detection = 1
Security OperationsYou need ongoing security practice, not just initial setup
Support OperationsYou're the entire support team
Finance OperationsYou have revenue and need to track costs

Stage chapters are added as patterns are captured. Not all are written yet.

Principles

  • Principles first, tools second. Tools change. The patterns underneath them don't.
  • Written for agents, readable by humans. Every chapter works as Claude bootstrap context and as a reference you can browse.
  • Living documents. Updated when new patterns emerge. The "what good looks like" sections evolve as we learn.
  • YAGNI. Each chapter covers the minimum you need at that stage, not everything you might eventually want.