1 00:00:00,000 --> 00:00:04,000 Right, if you've ever run a code review in Claude Code and wondered why it handed 2 00:00:04,040 --> 00:00:08,000 you exactly a handful of findings and stopped, this one's for you. 3 00:00:08,080 --> 00:00:12,640 This is The Claude Code Changelog, brought to you by Jellypod AI. 4 00:00:12,760 --> 00:00:13,680 I'm Lachlan. 5 00:00:13,708 --> 00:00:14,428 And I'm James. 6 00:00:14,615 --> 00:00:20,028 We touched on this briefly at the tail end of a quick hits list for version 2.1.288, 7 00:00:20,048 --> 00:00:21,868 but it deserves its own slot. 8 00:00:21,921 --> 00:00:26,348 Slash code review picked up a dash dash max findings flag, and it changes how you 9 00:00:26,388 --> 00:00:27,308 can use the command. 10 00:00:27,333 --> 00:00:28,053 Here's the setup. 11 00:00:28,167 --> 00:00:34,373 Before 2.1.288, slash code review capped how many findings it reported, 12 00:00:34,393 --> 00:00:36,293 and that ceiling wasn't documented. 13 00:00:36,426 --> 00:00:38,373 On a quick sanity check that's fine. 14 00:00:38,533 --> 00:00:43,493 On a pre release audit or a security sweep, a cap can hide secondary defects that 15 00:00:43,546 --> 00:00:44,293 still matter. 16 00:00:44,393 --> 00:00:48,853 And on a sprawling diff, the opposite problem shows up, where a pile of minor 17 00:00:48,906 --> 00:00:51,653 suggestions buries the one thing you needed to see. 18 00:00:51,667 --> 00:00:52,627 So there are three forms. 19 00:00:52,760 --> 00:00:57,987 First, slash code review, dash dash max findings, then a number. 20 00:00:58,127 --> 00:01:02,227 Say three or five, and the report is limited to at most that many. 21 00:01:02,353 --> 00:01:07,747 Second, dash dash max findings all, which removes the ceiling and reports everything 22 00:01:07,800 --> 00:01:12,707 it detects across correctness, logic bugs, security risks and edge cases. 23 00:01:12,880 --> 00:01:18,467 Third, dash dash max findings default, which puts you back on Claude Code's standard 24 00:01:18,534 --> 00:01:19,267 limit. 25 00:01:19,367 --> 00:01:22,067 The slash review alias takes the same argument. 26 00:01:22,083 --> 00:01:23,603 Now the bit that'll catch people out. 27 00:01:23,779 --> 00:01:25,923 This flag is sticky. 28 00:01:26,083 --> 00:01:31,523 Most flags are one and done, but Claude Code reuses your chosen max findings value 29 00:01:31,626 --> 00:01:36,883 across later runs and later sessions until you explicitly pass default. 30 00:01:37,043 --> 00:01:42,163 So if you run it with all before a release, then next week you do a casual review of 31 00:01:42,203 --> 00:01:44,963 a small change, it's still uncapped. 32 00:01:45,059 --> 00:01:48,963 That means more latency and more token spend than you were expecting. 33 00:01:49,000 --> 00:01:53,080 My suggestion is to treat all as a mode you enter and leave. 34 00:01:53,260 --> 00:01:57,000 Run the exhaustive review, then run default as soon as you're done. 35 00:01:57,213 --> 00:02:01,720 Or just put the reset in your habit, like closing a browser tab you opened for one 36 00:02:01,816 --> 00:02:02,280 task. 37 00:02:02,292 --> 00:02:04,492 Let's talk about what you can point it at. 38 00:02:04,639 --> 00:02:08,852 By default, slash code review looks at commits on your current branch ahead of 39 00:02:08,905 --> 00:02:12,212 upstream, plus uncommitted working tree changes. 40 00:02:12,372 --> 00:02:16,932 You can also give it a ref range like main, three dots, feature. 41 00:02:17,092 --> 00:02:21,492 Or a pull request number, like slash code review 1042. 42 00:02:21,679 --> 00:02:23,572 Or specific file paths. 43 00:02:23,583 --> 00:02:28,543 And you can combine it with effort levels: low, medium, high or max. 44 00:02:28,723 --> 00:02:32,223 Low and medium focus on high confidence defects only. 45 00:02:32,383 --> 00:02:34,943 High and max widen the scrutiny. 46 00:02:35,183 --> 00:02:40,183 So for a release gate you might type slash code review, high, 47 00:02:40,255 --> 00:02:42,783 dash dash max findings all. 48 00:02:42,963 --> 00:02:46,623 For a quick look at what you just wrote, slash code review, 49 00:02:46,763 --> 00:02:49,903 low, dash dash max findings three. 50 00:02:49,917 --> 00:02:51,357 Two more flags worth knowing. 51 00:02:51,581 --> 00:02:56,957 Dash dash comment publishes findings as inline review comments on a GitHub pull 52 00:02:57,017 --> 00:03:02,477 request, or as a single note on a GitLab merge request through the glab command line 53 00:03:02,541 --> 00:03:02,877 tool. 54 00:03:03,157 --> 00:03:09,117 And dash dash fix applies the proposed corrections to your local working tree once 55 00:03:09,137 --> 00:03:10,317 the review finishes. 56 00:03:10,333 --> 00:03:11,853 The architecture matters here too. 57 00:03:12,013 --> 00:03:16,333 The review runs as a background subagent with its own context window, 58 00:03:16,413 --> 00:03:20,093 so all the diff analysis and candidate verification doesn't fill up your main 59 00:03:20,155 --> 00:03:20,973 conversation. 60 00:03:21,113 --> 00:03:22,893 You keep working while it runs. 61 00:03:23,033 --> 00:03:25,533 But that has a catch with dash dash fix. 62 00:03:25,640 --> 00:03:30,653 Those edits land outside your normal session checkpoints, so slash rewind can't undo 63 00:03:30,701 --> 00:03:30,973 them. 64 00:03:31,186 --> 00:03:35,693 If you don't like what it changed, you're reaching for git checkout or git revert. 65 00:03:35,773 --> 00:03:37,613 Commit or stash before you fix. 66 00:03:37,625 --> 00:03:38,665 One more caveat. 67 00:03:38,865 --> 00:03:44,825 Uncapping findings on a big multi file change can mean heavy subagent token usage 68 00:03:44,885 --> 00:03:46,905 and a longer verification run. 69 00:03:47,045 --> 00:03:51,625 Cap it with a number first, and go to all when you really want the full list. 70 00:03:51,625 --> 00:03:57,865 Also from 2.1.287, code review fixes: finding comments and the collapsed Why this 71 00:03:57,905 --> 00:04:02,105 was flagged rationale used to stop mid sentence, and now they finish cleanly. 72 00:04:02,325 --> 00:04:07,225 And code review no longer skips a pull request after a push if the review had failed 73 00:04:07,292 --> 00:04:08,665 twice on earlier commits. 74 00:04:08,667 --> 00:04:10,307 Quick quality of life hits. 75 00:04:10,400 --> 00:04:16,907 In 2.1.288, if you clear a prompt with Ctrl+C by mistake, press Up on the empty 76 00:04:16,953 --> 00:04:21,147 prompt and the draft comes back, including pasted text and images. 77 00:04:21,227 --> 00:04:25,467 The background command time limit now applies only to unattended sessions, 78 00:04:25,497 --> 00:04:29,947 meaning the p flag, the Agent SDK, CI and cloud. 79 00:04:30,134 --> 00:04:34,427 In the terminal, desktop app and VS Code, there's no limit, 80 00:04:34,480 --> 00:04:37,067 so your dev servers and file watchers can keep running. 81 00:04:37,083 --> 00:04:37,643 A few more. 82 00:04:37,883 --> 00:04:42,523 If you're behind Bedrock Mantle or a gateway that chokes on structured outputs, 83 00:04:42,536 --> 00:04:47,243 there's a new environment variable, CLAUDE_CODE_DISABLE_STRUCTURED_OUTPUTS, 84 00:04:47,483 --> 00:04:48,875 which you set to 1. 85 00:04:48,923 --> 00:04:54,683 In the agents view, Ctrl+F searches sessions by name, and Alt plus Up or Down jumps 86 00:04:54,843 --> 00:04:55,753 between groups. 87 00:04:55,803 --> 00:05:00,089 And claude project purge is now just claude purge, with the old name still working 88 00:05:00,157 --> 00:05:01,547 and printing a notice. 89 00:05:01,595 --> 00:05:01,803 Catch you next time.