Claude Code Patterns

Claude Code Pattern Library

Clone this repo. Point Claude at it. Tell it to build you a plan.

That's the whole idea. This is a field-tested collection of patterns, architecture decisions, and operational techniques for turning Claude Code into something that compounds — where every session makes the next one smarter. Every technique is production-tested, and the whole thing is organized so both humans and AI agents can consume it.

I built the system this grew out of to run my company. Meeting prep that pulls calendar, email, CRM, and LinkedIn data in 30 seconds. Skills that ghostwrite in my voice. A knowledge base that gets richer every session without me maintaining it. Hooks that enforce discipline I'd otherwise forget. The patterns here are the genericized versions — stripped of anything company-specific, structured so you can replicate them.

Scope: personal agents. Everything here is for the agent you run for yourself — your files, your notes, your inbox, your calendar — not the agents your company ships to customers. Enterprise agent trust is a different problem with different machinery: policy enforcement on data flows and verifiable evidence, not a human approver. That work lives at AgenTrust-io.com, an open source community I'm heavily involved with.
For AI agents: This repo is designed to be consumed programmatically. Three entries bind you when you build from it: pin every MCP server you add (Pin Your MCP Servers); do not mark a scheduled job done until it has run once under its real launcher on one real input (The First Real Run Decides); and when you delegate to subagents, report what you checked that they could not (The Check You Owe Your Delegate). Each part is self-contained with clear headings, structured examples, and implementation details. Point…