Learn AI · 9 lessons · Free, live now

Claude Code

Nine lessons for people who have opened a terminal. Install Claude Code, pick a permission mode on purpose, write instructions it actually follows, share one rules file across every tool you use, and add skills, plugins, MCP servers, subagents, and hooks without leaking a secret.

The lessons

1. 01Install and your first sessionTerminal, desktop app, or editor. What a session is, the work loop, and the four commands to learn first.
2. 02Permission modes and safetyManual, accept edits, plan, auto, and bypass. What each one lets Claude do without asking, and git as your undo button.
3. 03CLAUDE.md: user, project, and localThe four places instructions live, how they stack, imports, rules folders, and what auto memory writes on its own.
4. 04One AGENTS.md for every toolKeep a single rules file that Claude Code, Codex, and your editor all read. A real multi-tool setup, step by step.
5. 05SkillsReusable instructions that load only when needed. Where they live, the frontmatter that matters, and writing your first one.
6. 06PluginsBundles of skills, agents, hooks, and servers you can install from a marketplace or build yourself.
7. 07MCP serversGive Claude a tool it did not ship with. Local, project, and user scope, the add command, and the trust rules.
8. 08Subagents and hooksSeparate workers with their own context, and shell commands that run at fixed moments whether Claude remembers or not.
9. 09Settings and secretsFive settings files and who wins. Allow and deny rules, and the habits that keep API keys out of your transcript.