Build with Claude Code: Hooks

Automate what Claude Code feeds itself, and stop fighting the same battles twice.

Hooks are shell scripts that run automatically at specific moments in Claude Code’s loop — right before it runs a tool, right after it edits a file, when you submit a prompt, when a session ends. They let you bake your own rules into the workflow so Claude doesn’t have to remember them and you don’t have to enforce them by hand. This is a follow-up to the Basics session. It assumes you’ve already built something with Claude Code — if you’ve felt the pain of "I keep telling Claude the same thing" or "my tests are always stale after an edit," you’ll know exactly why hooks matter. If you’ve never written a shell script, you’ll still be fine — we use tiny ones and explain each line. By the end of the 90 minutes you will: • Know every hook event (PreToolUse, PostToolUse, UserPromptSubmit, SessionStart, and the rest) and when each one fires. • Have written your own hooks for your own project — auto-format on edit, run tests before commits, inject context into every prompt. • Know how to debug a hook that isn’t firing or is doing the wrong thing. • Know when a hook is the right tool and when a skill or CLAUDE.md instruction is the better choice. How the session runs: 15 minutes on what a hook actually is and the event lifecycle. Then we install a real hook together — auto-format after every edit. Next we build one that injects your current git branch into every prompt. Back half: you pick a pain point in your own project and write a hook for it, with live help. Bring your laptop, charged, and a project you’re actively working on. We’ll write hooks for that project, not toy examples.