You have been using Claude Code for weeks, maybe months. You know the basics — ask it to write code, review a PR, refactor a function. But somewhere around hour three of a long session, things start to break down. Claude forgets the coding standards you told it twenty minutes ago. Your context window is bloated with exploration output that should have happened somewhere else. You have a CLAUDE.md file with 400 lines of instructions, half of which are irrelevant to what you are doing right now.
The problem is not Claude Code. The problem is that you are treating a configurable engineering system like a chatbot. Claude Code has a layered customisation architecture — persistent context, on-demand knowledge, isolated execution, and deterministic guardrails — but most developers use exactly one layer: stuffing everything into CLAUDE.md and hoping for the best.
This course teaches you the full system. Not theory. Not feature overviews. The exact file structures, configuration patterns, and workflow designs that separate developers who fight their AI assistant from developers who have engineered it into the most productive member of their team.
What you'll produce
- A clear understanding of Claude Code's six-layer customisation hierarchy and when each layer applies
- A properly structured CLAUDE.md with imports, rules, and path-specific instructions that load only when relevant
- Reusable skills for your most common workflows — PR review, deployment, changelog generation, code migration
- Isolated agents that handle heavy tasks in their own context window without polluting your conversation
- The compounding pattern: agents preloaded with skills that produce consistent, standards-compliant output from the first run
- Deterministic hooks that enforce rules Claude cannot ignore — linting, security checks, formatting, audit logging
- A complete project configuration blueprint you can drop into any repository
Who this is for
- Individual developers using Claude Code daily who want to stop repeating themselves and start building reusable workflows
- Tech leads and architects who need Claude Code to follow team standards consistently across every developer's session
- DevOps and platform engineers building internal tooling around Claude Code for their organisation
- Anyone who has hit the wall with a bloated CLAUDE.md and knows there must be a better way
Prerequisites
You should have Claude Code installed and have used it for at least a few sessions. Familiarity with markdown, YAML frontmatter, and basic command-line tools is assumed. No prior experience with skills, agents, or hooks is needed — that is what this course teaches.
Time commitment
~3 hours of interactive learning. Best done with your actual project open alongside the course, building configuration files as you go.
Course modules
- Why Your Configuration Matters — The problems that surface when you treat Claude Code like a chatbot instead of a configurable system
- The Customisation Hierarchy — Six layers, from permanent context to deterministic enforcement, and how they compose
- CLAUDE.md and Rules — Your project's permanent memory: file locations, imports, path-specific rules, and what belongs here versus elsewhere
- Building Skills — On-demand recipes: file format, triggering, shell injection, string substitution, and practical templates
- Building Agents — Isolated specialists with their own context window, tool restrictions, worktree isolation, and persistent memory
- Skills Inside Agents — The compounding pattern that changes everything: four workflow designs for real engineering work
- Hooks and Guardrails — Deterministic automation: enforce rules Claude cannot ignore, at every lifecycle point
- Your Configuration Blueprint — The decision framework, a starter project structure, and the path from first skill to full team adoption
Ready? Start with Why Your Configuration Matters