Tools for people who build with AI.
Direct. Technical. Opinionated.
// featured product
A complete operating system for writing prompts that work in production.
Covers everything from first principles to multi-agent orchestration — with the templates and code to skip straight to applying it. Built from real production usage of Claude Code, not conference talks. 30 chapters, 8 template schemas, 100+ templates, 125 code examples.
// all products
Each product starts as something we needed and couldn't find. No fluff added to hit a word count.
14 project-specific CLAUDE.md templates. Drop one into any repo and your Claude Code session inherits the right conventions immediately. Python, TypeScript, Next.js, Go, and more.
6 printable reference cards covering Claude Code commands, prompt patterns, agent architectures, tool-use patterns, CLAUDE.md syntax, and hook triggers.
Build custom Claude Code tools without starting from scratch. Includes server scaffold, tool registration patterns, error handling, and 4 example servers ready to fork.
Systematic prompt evaluation methodology with scoring rubrics, regression test harnesses, and A/B comparison tooling. Stop guessing if your prompt got better.
Vault template and Claude config wired together out of the box. Pre-built folder structure, Dataview queries, Templater setup, and Claude memory hooks. Skip the 20-hour yak shave.
22 ready-to-use automation hooks for Claude Code. Pre-task validation, post-task reporting, damage control, linting enforcement, session logging, and more.
16 chapters on multi-agent patterns from production. Builder/validator loops, tool-use architectures, memory systems, rate-limit handling — everything you'd learn the hard way.
AI-assisted code review patterns that actually catch things. Prompt templates for security, architecture, and correctness reviews. Includes CI integration examples.
One-page command reference covering every slash command, useful flags, keyboard shortcuts, and hook trigger events. Printable. No email required.
// about
Direct, technical, opinionated. We build tools for people who are already using AI to build things — not introductory content for people wondering if they should try it.
Every product starts as something we needed and couldn't find. The Prompt Engineering OS came from months of running Claude Code in real projects. The patterns in it are distilled from what actually worked, not what sounded good in a tweet thread.
The backlog is full of tools that follow the same principle: specific, opinionated, immediately useful.