Hooks on Bruce on AI Engineering

MCP vs Skills vs Hooks in Claude Code: Which Extension Do You Need?

Thu, 02 Apr 2026 10:00:00 +0800

Claude Code has three distinct extension mechanisms: MCP (Model Context Protocol), Skills, and Hooks. They look related on the surface, but they operate at fundamentally different layers of the system. Choosing the wrong one means wasted effort, bloated context windows, or brittle automation that breaks when you need it most.

This guide provides a deep technical comparison — how each extension works internally, what it costs in terms of tokens and complexity, and a clear decision framework for when to use which.

Claude Code Hooks: Automate Your AI Workflow (2026)

Wed, 04 Mar 2026 09:00:00 +0800

Claude Code is probabilistic by nature. Ask it to format your code, and it might. Ask it to never touch .env files, and it usually respects that – until it doesn’t.

Claude Code hooks solve this. They are shell commands, HTTP endpoints, or LLM prompts that run automatically at specific points during Claude’s operation. Before a file edit happens, after a command runs, when a session starts, when Claude finishes a task. Hooks give you deterministic control over the parts of your workflow that cannot be left to chance.

Claude Code Hooks 2026: All 9 Events + 12 Copy-Paste Examples

Thu, 26 Feb 2026 10:00:00 +0800

Claude Code is probabilistic. You give it the same prompt twice and get different results. That’s fine for creative work — but your engineering workflow needs deterministic guarantees.

You need files auto-formatted on every save. You need .env files locked down. You need rm -rf / blocked before it ever reaches your shell. You need these things to happen every single time, not just when the AI remembers.

Claude Code Hooks Guide: 12 Ready-to-Use Configs for Automation

Wed, 18 Feb 2026 11:00:00 +0800

If you use Claude Code daily, you have probably run into these problems:

Claude edits a file and the formatting is a mess – you have to manually run prettier every time
Claude accidentally modifies .env or package-lock.json, and you only notice when it is too late
Claude says “done!” but the tests were never run
You step away to grab coffee and have no idea whether Claude is waiting for your input

The root cause is simple: Claude is probabilistic, but your workflow needs deterministic guarantees.