On May 21, 2026, Claude Code 2.1.147 renamed /simplify to /code-review. The rename is the small part. The substantive change is in the body of the release note: “The old cleanup-and-fix behavior has been removed.”
If you used /simplify to make Claude apply changes, that path is gone. /code-review reads the diff, reports correctness bugs at the effort level you pass, and stops there. You apply the fixes yourself.
The New Command Shape
/code-review high
/code-review medium
/code-review low
/code-review high --comment
The effort flag picks how hard the model thinks before flagging issues. The --comment flag posts findings as inline GitHub PR comments — which is the genuinely new capability in this release, not the rename.
Until yesterday, getting Claude to leave inline comments on a PR meant a plugin, a custom skill, or piping output into gh pr review. /code-review --comment collapses that into one invocation.
Why The Behavior Split Matters
/simplify did two things in one pass: identify issues and fix them. That worked when the changes were obvious. It failed badly when the model decided a five-line tweak should also be a 200-line refactor. The split into “review-only with effort knob” + “you apply” is a more honest separation of concerns.
The cost is that the previous one-shot workflow is now two steps. If you had a skill, a saved prompt, or a habit built around /simplify landing changes, you need to do the apply step yourself — either by hand or by chaining a second command after the review.
The Bash 127 Regression
There is one more thing worth knowing if you updated yesterday. Version 2.1.147 introduced a Bash tool regression where some users got exit code 127 on every shell command. Anthropic shipped 2.1.148 about five hours later — published 2026-05-22 at 01:16 UTC — with the single-line fix.
If you ran claude --version this morning and saw 2.1.147, and your Bash tool calls are failing with code 127, update.
npm install -g @anthropic-ai/claude-code@latest
What To Update In Your Setup
A short checklist for anyone running Claude Code in something other than a vanilla terminal:
- Replace
/simplifywith/code-review <effort>in any saved prompts, skills, hooks, or shell aliases. - Decide whether you want the auto-apply behavior back. If yes, add an explicit apply step after the review.
- If you review GitHub PRs through Claude Code, look at
--comment— it likely shortens your existing flow. - If your Bash tool started failing yesterday, jump to 2.1.148.
The Pattern
Claude Code’s slash-command surface is being split apart so that each command does one thing. /simplify was a convenient bundle. /code-review is the more honest primitive. The matching change on the SDK side last week — splitting subscription billing from Agent SDK billing — points in the same direction: features that used to be implicit are being made explicit, and the cost of relying on the implicit version is going up.
Pin the Claude Code version in any setup where you depend on a specific command behavior. The release notes are short, which is exactly why these changes are easy to miss in a hurry.
Part of the Claude Code series
- Claude Code v2.1.100+ Burns ~20K Phantom Tokens Per Request — the prior small-change-big-bill story
- Third-Party Claude Agents Lose the Subscription Subsidy June 15 — the billing-side split
Sources: