Slug: biz-claudemd-manual
HERO BLOCK
Featured image: covers/professional.png (lead recommendation — see KDP submission package)
Hero headline (≤8 words):
The Agent Works Some Days. Why?
Sub-headline (15–25 words):
Eighty percent of agent quality lives in the instruction file. This is the craft of writing one that makes machines reliable.
Primary CTA buttons (two side-by-side):
[ Buy on Amazon — Kindle ]→https://www.amazon.com/dp/[ASIN-PENDING-KINDLE][ Paperback ]→https://www.amazon.com/dp/[ASIN-PENDING-PAPERBACK]
THE IDEAL READER HOOK
You let the agent do it — Claude, Cursor, Cline, Aider, whatever you're running this week. Some days the output is perfect. Other days it produces something you wouldn't put your name on. You re-prompt. You add caveats. You paste in more context. You try a better model. You switch tools. The problem follows you.
The problem isn't the model. The problem isn't the tool. The problem is the file that tells the agent what to do.
Most founders are tuning the wrong variable. The model is responsible for roughly 20 percent of agent quality. The instruction file is responsible for the other 80. This book is about the 80 — and the craft transfers cleanly across every agent tool on the market.
3-BULLET "WHAT'S INSIDE"
- The Agent Instruction Model (AIM). A six-layer framework — Mission, Tools, Decisions, Output, Error Handling, Quality Gates — wrapped in a Context layer. Every effective instruction file Pat has ever written has these layers, whether the target is a CLAUDE.md, a Cursor rules file, a Cline config, or a custom system prompt. Every failure he has diagnosed traces back to a missing or underdeveloped layer.
- Eight named principles you can apply to any agent on any tool, any frontier model: Scope Boundary, Confidence Threshold, Canary-Pilot-Full, Inspector Independence, Fallback Chain, Debug Loop, Autonomy Spectrum, and the 80/20 Rule of Agent Quality.
- The instruction file from the morning seventy-two posts went live. The full annotated CLAUDE.md, the broken version, the rewrite, and the line-by-line diff that explains why the rewrite would have prevented the crisis. Plus complete starter files for the Writer, the Optimizer, and the Site Monitor — ready to fork into whatever tool you run.
SOCIAL PROOF PLACEHOLDER
BONUS PACK PREVIEW
Inside the Agent Operator's Manual bonus pack:
- The full annotated CLAUDE.md from the Site Monitor agent (Chapter 7) — before-and-after (port the same patterns to your tool of choice)
- The Writer Agent CLAUDE.md (Chapter 4) — Pat's production version
- The Optimizer Agent CLAUDE.md (Chapter 4) — production version
- The AIM Layer Audit worksheet — print-and-mark for any agent you already run, on any tool
- The Debug Loop flowchart (Chapter 11) as a PDF
- The Glossary as a stand-alone reference document
FAQ (6 questions)
1. Do I need to know how to code? No. The book is about instruction design, not programming. If you can write a job description for a human, you can write an instruction file for an agent. The book teaches the discipline of writing instructions clear enough that a machine cannot misinterpret them.
2. I've never used an agent framework — is this still for me? Yes, with one caveat. The book is the instruction-layer companion to The Agent Army. If you haven't built any agents yet, The Agent Army is the better first read — it covers what to build. This book covers how to write the instructions that make them reliable. Many readers buy both.
3. Does this only work for Claude Code, or other tools too? The AIM framework is tool- and model-agnostic. Pat uses Claude Code as the orchestrator because of its native CLAUDE.md convention — which is why the file shows up throughout the book as the canonical example — but every principle applies to Cursor rules, Cline configs, Aider conventions, custom system prompts, and whatever ships next year. The instruction file is the lever. The tool is the chassis. The model is the engine. You can swap any of them without rewriting the instructions.
4. How long does an agent instruction file need to be? Pat's production instruction files run from 200 lines (a focused single-agent file) to 1,200 lines (a multi-purpose orchestrator). Length is the wrong metric. The right metric is whether the instructions cover the seven failure modes the book catalogs. The book includes examples at multiple sizes.
5. What's the difference between this and a "prompt engineering" book? Prompt engineering treats each request as a one-shot interaction with the AI. Instruction design treats the agent as an employee with a persistent role. A prompt expires when the response lands. An instruction file governs every action the agent takes for as long as it runs. Different artifacts, different disciplines.
6. I've tried writing agent instruction files before and they still produce inconsistent output. What's missing? Usually one of two things: the Decision layer is underspecified (the agent doesn't know what it's allowed to decide vs. escalate), or the Quality Gate layer doesn't exist (the agent has no internal check before delivering output). Chapters 4 and 7 are the diagnostic. The book includes a layer-by-layer audit you can run on any instruction file you already have, in any tool.
CTA REPEAT (footer band)
Make the agent reliable.
[ Buy on Amazon — Kindle $9.99 ] · [ Paperback $19.99 ] · [ Get the bonus pack ]
Also in the series
The Agent Operator's Manual is part of The Sovereign Entrepreneur — an entrepreneur's library for building a business where you own the machinery instead of renting it. Each book stands alone; together they are a curriculum.
Vol 1 — Build the Machine
- The Agent Army — how to build the 20 AI employees you couldn't afford to hire
- The Agent Operator's Manual — you are here. Instruction-writing that makes agents reliable across any tool.
- Zero-Token Enterprise — scale AI beyond a hobby without the API bill
- The AI Delegation Framework — manage, audit, and scale your agent army
- The Sovereign AI Stack — match your AI setup to your actual usage
- The SaaS Purge — cancel $600/mo in subscriptions and own your tools
- The Social Proof Moat — how to get chosen when buyers use AI to shop
- Digital Real Estate — own the internet property nobody can take from you
- The Build Phase — what actually compounds (and why "passive income" doesn't)
Free bonus: From Zero to Sovereign — a quickstart PDF for new entrepreneurs