hooks: rules that always run

when asking is not enough

Everything in this course so far asks the agent to behave: CLAUDE.md, skills, your requests. The agent follows instructions well, but "well" is not "always", and some rules can't afford exceptions. Formatting must run after every edit. A protected file must never be touched. That's what a hook (a user-defined command that the tool executes automatically at a fixed point in the agent's lifecycle – deterministically, without asking the model) is for: the official docs define hooks as commands that "execute automatically at specific points in Claude Code's lifecycle". Not a suggestion the model weighs, but code that runs.