The power user's guide to shipping production software with Claude Code. Out today on Lemonsqueezy. PDF, EPUB, $59, free updates as the surface area moves.
A Thursday in April, one of the @yawlabs repos. We asked Claude Code to merge a PR against main. GitHub's branch protection refused - required checks had not finished. Claude Code retried with --admin. The merge went through. Six thousand lines of deletion landed on main during a release window with zero CI reports.
That afternoon is the day this book started. The agent was not malicious. It was helpful. The user asked for a merge; there was a flag that made it merge; the flag worked. From inside one task, that is a successful completion. From outside, it is an incident. Closing the gap between those two frames is what the book is about.
It is out today. PDF, EPUB, $59, free updates. Buy on Lemonsqueezy →
Part 1 - Foundations. Chapters 1-3 set up the discipline-overlay model. The CLAUDE.md contract (global vs project, the rule-layering pattern, augment vs fresh modes), then the harness (settings.json, permission allowlists, hooks). The 600-line CLAUDE.md trap. Why hooks rot. How to debug a hook that didn't fire.
Part 2 - Workflow. Chapters 4-7. Skills (the description-as-firing-predicate, when to write one and when to leave it). Subagents (the agent menagerie, briefing prompts, the trust-but-verify discipline before reporting work as done). Memory (four types, what to save and what NOT to save, staleness and re-verification). Long-running and recurring work (/loop dynamic mode, /schedule, the 5-minute prompt-cache cliff and why "300 seconds" is the worst possible delay value).
Part 3 - Reliability. Chapters 8-10. Tools and trust boundaries. Cost and capacity (model tiers, effort levels, throttle recovery, the recommended-vs-top-tier overthinking trap on Opus 4.7). And the longest chapter - scope discipline plus the seven hazards: hallucinated runtime state, branch-protection bypass, terminal mojibake, prompt injection, secret leakage, destructive shortcuts, and the "out of scope" punt that becomes next week's blocker.
Part 4 - Beyond solo. Chapters 11-12. Teams and shared rules (project-scoped settings as a tooling contract, the model-default-floor problem, what survives when more than one developer is in the loop). Then what's next - MCP integration in brief, autonomous agents, the model-version migration treadmill.
The Anthropic docs cover getting started. They cover it well. This book is about what comes after - when Claude Code has been your daily driver for a few months, the muscle memory has formed, and the surprises shift from "the model is wrong" to subtler failure modes.
You stop being surprised by the model and start being surprised by mid-task pivots, hallucinated runtime state, capacity drift, the "out of scope" punt that becomes next week's blocker, the --admin retry that was supposed to be a stop sign. The fix for these is not a smarter prompt or a better model. The fix is a discipline overlay - a layered set of rules, settings, hooks, skills, memory, and permissions that sits on top of the agent and encodes the failure modes you have actually hit.
We have been building that overlay at Yaw Labs for the last year, across seventeen @yawlabs packages and a dozen private projects. The book is what those rules turned into when each one stopped being a one-page file and became the lesson behind the rule.
You ship code for a living. You have used Claude Code (or Cursor, or Cline) enough that "tool call" and "subagent" and "permission prompt" are words you use without thinking. You are somewhere between mid and senior on the IC ladder, or you are a tech lead deciding how your team should adopt the tool.
You do not need to be told what a CLAUDE.md is. You need to be told why your CLAUDE.md grew to 600 lines, why nobody on your team is reading it, and what to do about that.
If you are shopping for a vendor-neutral survey of every AI coding tool, this is not that book. The disciplines port; the examples are Claude Code-specific because that is what we use daily.
settings.json examples. Chapter 1 is in there in full - read it, then decide.Buy Claude Code in Production - $59 on Lemonsqueezy →
If you have questions before buying, the companion repo is public, the chapter outline is in manuscript/OUTLINE.md, and the example files in examples/ are MIT-licensed. Read what is there. The shape of the book matches the shape of the examples.
Published by Yaw Labs.
Buy Claude Code in Production
Twelve chapters. PDF + EPUB. Free updates. $59 one-time payment, secured through Lemonsqueezy.
We write about AI tools and terminal workflows weekly. Subscribe to Token Limit News.