---
title: "BuilderBooks"
description: "Curated business books as agent skills for Claude Code, Codex and Lovable. Give AI the frameworks directly, packaged as SKILL.md."
url: "https://builderbooks.dev/"
type: "page"
site: "BuilderBooks.dev"
---

# Books for people who build with AI.

Stop being the middle man, who is retelling AI the books you read. Instead give the frameworks and best practices from your favorite books, packaged as `SKILL.md`, directly to your agents. No more lost in translation.

Works with Claude Code, Cursor, ChatGPT and Claude Projects. See [/install](/install).

## What the heck are skills?

A skill is context you load into your AI agent before you start a task. It is a single markdown file that tells the agent what you value, how you decide, and which framework you trust, so the agent does not have to guess.

We did not make the format up. `SKILL.md` was introduced by Anthropic in October 2025 to teach Claude Code about a project, a workflow, or a taste. In December 2025 Anthropic published it as an open, cross-platform standard at <https://agentskills.io>. We package each book as a SKILL.md so the framework reaches your agent directly.

## Stop being the Middleman

Before, you read three hundred pages, kept a fraction, and retold that fraction to your agent. Two rounds of loss, and the second one is invisible, because you pass on what you remembered, not the framework.

Now the framework reaches your agent directly. You still decide which book matters. You just stop being the middleman.

## Why not simply give the agent the book as PDF?

You can try, but most books are not a clean PDF. They ship as EPUB, Kindle or DRM locked files that agents cannot read reliably. Even with a clean file, a book is 60,000 words of story, anecdotes and repetition: the agent has to guess what matters and your context window fills up with noise. A skill is the framework only, a few hundred lines written as rules, steps and edge cases, so the agent acts instead of paraphrasing.

## Three steps

1. Pick a skill
2. Drop it in `~/.claude/skills/`
3. Run a command

Setup for Claude Code, Cursor, ChatGPT and more: [/install](/install).

## What's the actual problem?

| Cluster | The problem | Book | How it loads | Outcome |
| --- | --- | --- | --- | --- |
| Activation | They sign up and never come back. | Product-Led Onboarding, Ramli John | Run `/activation-audit` | An annotated funnel with the exact step where people drop off |
| Discovery | Everyone says they'd use it. Nobody does. | The Mom Test, Rob Fitzpatrick | Run `/interview-guide` | An interview script that rules out the questions people politely say yes to |
| Design | It works, but it looks like every other Lovable template. | Refactoring UI, Wathan & Schoger | Loads while you build | Concrete rules on hierarchy, spacing and weight, applied to your components |
| Positioning | I can't explain what I built in one sentence. | Obviously Awesome, April Dunford | Run `/positioning-canvas` | A position that holds up against the alternative your customer actually considers |
| Pricing | I have no idea what to charge. | $100M Offers, Alex Hormozi | Run `/value-equation` | An offer the price no longer has to justify |
| Go-to-Market | I posted it once. Now what? | Traction, Weinberg & Mares | Run `/bullseye` | Three channels to test instead of nineteen to think about |

## Don't know where to start? Pick a problem.

Each bundle is the set of books for one of those moments. Every third skill is on us. See [/bundles](/bundles) and [/pricing](/pricing).

## We built this for our own product first.

People signed up for MeetMeAt and never came back. Instead of retelling our agent the onboarding book we had read, we handed it the framework as a skill and let it audit the real flow. That first file is why this library exists. Read the story at [/about](/about).