Automating Code Review with Claude Code Hooks

Every time I edited a file, I’d forget to review it properly. So I stopped doing it manually.

The problem

Code review is the right habit. Running it consistently is the hard part. I’d edit 10 files in a session and review 3.

The solution

Claude Code’s hook system lets you run arbitrary commands after tool use events. I wired a review step into PostToolUse:Edit.

Setup

In .claude/settings.json:

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Edit",
        "hooks": [
          {
            "type": "command",
            "command": "echo 'File edited, review queued'"
          }
        ]
      }
    ]
  }
}

Then in CLAUDE.md, I added instructions that trigger after each edit:

After editing any file, briefly check for: obvious bugs, missing error handling,
security issues. One sentence per concern. Skip if none.

Result

Every file edit now gets a lightweight automatic review pass. Claude flags issues inline as it works, not after the fact.

The point is timing: issues get flagged while the file is still open, instead of in a review pass I might skip.

Next step

Hooks can also backfire: a hook that rewrites commands fails every command when its binary is missing. For a sturdier pattern, see session goals with Stop hooks. Want review gates wired into your team’s Claude Code setup? That is part of the Claude Code workflows I build. Book a call.