Hooks and Settings — wiring external behavior into tool execution
There are places where you want external behavior to run before or after an LLM agent's tool call. Auto-formatter execution, change watching, logging, policy validation, secret-leak blocking — that kind of work.