Pass –max-findings on local /code-review, then reset it

At a glance

Claude Corner tip
  • Claude Code v2.1.288 and later accept --max-findings on the local /code-review command, with a number N, all, or default.
  • The value is sticky: later reviews reuse it until you pass default, so reset when you are done.
  • On the morning of 7 October 2026, npm latest and next were v2.1.292 and had the flag, while stable was still v2.1.285 and did not, so leave stable fleets alone until policy moves them.

/code-review reviews a diff in your terminal without the GitHub App. With no target it covers your branch’s commits ahead of upstream plus any uncommitted changes. The review stops at a usual findings limit, and the docs do not publish that number. On a large diff you cannot tell whether Claude stopped because the branch is clean or because the cap filled. It shipped in v2.1.288 on Friday, 2 October, at 4:19 PM ET. Our Saturday brief, GitHub Retires Four Copilot Models While Claude Code Stable Lacks Fail-Closed Fixes, gave it one sentence. This is the full how-to.

Why it matters

A cap that is too low hides real bugs on a big branch, and a cap that is too high buries the bugs that matter under cleanup notes. The flag lets you pick per job. The catch is that it sticks. The docs say later reviews reuse the value you typed until you pass default, so a number you set last week still shapes today’s review. Make resetting a habit.

How to set and reset the findings cap

This is the local terminal command. It is not the GitHub App Code Review on pull requests and not the ultra cloud review. The docs do not mention Claude Desktop, so do not assume it accepts the flag.

  1. Run claude --version and confirm it prints v2.1.288 or later. On the morning of 7 October 2026, npm latest and next are v2.1.292 and stable is v2.1.285, which lacks the flag.
  2. To move a native install to the latest channel, set autoUpdatesChannel to latest in settings.json (latest is the default unless your settings pin stable), then run claude update. stable typically runs about one week behind. For an npm install, install the latest tag as in the block below.
  3. In a repository with commits ahead of upstream or uncommitted changes, run a capped review such as /code-review --max-findings 12.
  4. When you want every finding on one review, run /code-review --max-findings all.
  5. When you are done, run /code-review --max-findings default to restore the usual limit for later reviews.
  6. /review is an alias of /code-review, so the same flag works there. To review something other than the default diff, add a target (a file path, a PR number, a branch name, or a ref range such as main...my-feature) on the same command.
bash Tutorial
# Check your build; you need v2.1.288 or later
claude --version
# npm installs that follow latest only (leave stable fleets alone):
# npm install -g @anthropic-ai/claude-code@latest
# Native installs with autoUpdatesChannel set to latest:
# claude update
▸ Show full code (6 lines)
# Check your build; you need v2.1.288 or later
claude --version
# npm installs that follow latest only (leave stable fleets alone):
# npm install -g @anthropic-ai/claude-code@latest
# Native installs with autoUpdatesChannel set to latest:
# claude update

To confirm, run a capped review, then a plain /code-review on a diff with plenty to report. Per the docs it reuses your number. After default, the next plain review returns to the usual limit. We did not run a signed-in review this morning, so treat that as a test of your own.

Gotchas

  • Sticky value: a number or all carries into later reviews until you pass default. Reset before you hand a repo to a teammate who expects the usual output.
  • All is wide: all on a large diff can bury correctness bugs under cleanup notes. Use it on purpose, not as a standing setting.
  • No published default: 12 is an example, not a step from a documented baseline.
  • No status line: nothing documented prints the stored cap, so the version check and the reuse behavior are your confirmation.
  • Other flags are separate: --fix applies findings to your working tree and --comment posts them to a pull request. Leave them off unless you mean it.
  • Stable fleets: npm stable was v2.1.285 on the morning of 7 October 2026 and lacked the flag. Leave those machines alone until your policy moves the channel.

Recommended AI prompt

Copy this paragraph into ChatGPT, Claude, Gemini, Grok, or whatever AI you use.

I use the local /code-review command in the Claude Code terminal and want to control how many findings it reports. Ask me for the output of claude --version first, and if it is older than v2.1.288, tell me the flag is missing and how to move to the latest channel, noting that on 7 October 2026 npm stable is still v2.1.285 while latest is v2.1.292. If I am on v2.1.288 or later, give me a short checklist for running a capped review with --max-findings and a number N, a full review with --max-findings all, and a reset with --max-findings default, and remind me that the value I pass is reused by later reviews until I reset it and that /review is an alias. Treat this as the terminal command only, not the GitHub App review or the cloud ultra review, and do not invent flags.

Go deeper in Grok

Explore each Top Story in Grok. Links open in a new tab. On phones, the same link may open the Grok app if you have it installed (via your device's normal link handling).

Article: Pass –max-findings on local /code-review, then reset it

Privacy: links open grok.com in your session only. AIDevPulse does not run your prompts through our API.

1 thought on “Pass –max-findings on local /code-review, then reset it”

Leave a Comment