1 00:00:00,260 --> 00:00:02,960 Quick heads up before your nightly pipeline bites you. 2 00:00:03,520 --> 00:00:07,720 This is The Claude Code Changelog, brought to you by Jellypod AI. 3 00:00:08,340 --> 00:00:10,440 I'm Lachlan, and with me is James. 4 00:00:11,093 --> 00:00:11,693 Thanks, Lachlan. 5 00:00:12,293 --> 00:00:18,273 Today is one change from Claude Code 2.1.309, which shipped on October 22nd. 6 00:00:18,993 --> 00:00:23,253 The claude ultrareview command now requires workspace trust. 7 00:00:23,713 --> 00:00:25,373 In a terminal, it asks you to confirm. 8 00:00:25,873 --> 00:00:27,413 Anywhere else, it refuses to run. 9 00:00:27,953 --> 00:00:32,393 If you've got ultrareview in a script or a CI job, it will fail after you update, 10 00:00:32,693 --> 00:00:33,413 until you change it. 11 00:00:34,032 --> 00:00:36,292 Let's cover why it's built that way. 12 00:00:37,212 --> 00:00:39,672 Ultrareview isn't the local code review. 13 00:00:40,052 --> 00:00:44,252 It's a deep review that runs as a cloud session on Anthropic's infrastructure, 14 00:00:44,612 --> 00:00:48,252 with a fleet of reviewer agents going over your branch or pull request. 15 00:00:49,112 --> 00:00:52,932 So your repo state leaves your machine and gets worked on in the cloud. 16 00:00:53,792 --> 00:00:57,532 The reasoning in the brief is that if the workspace is untrusted, 17 00:00:58,052 --> 00:01:04,012 say a cloned open source repo or an unvetted PR, then config files or prompt 18 00:01:04,072 --> 00:01:07,452 injection payloads in that code could mess with the review. 19 00:01:08,632 --> 00:01:12,472 That part is interpretation of the design, not something the release notes spell 20 00:01:12,572 --> 00:01:12,792 out. 21 00:01:13,572 --> 00:01:18,892 What the notes do say is plain: it asks on a terminal, otherwise it refuses. 22 00:01:19,590 --> 00:01:21,390 So here's how you actually deal with it. 23 00:01:21,810 --> 00:01:23,110 Interactive use is easy. 24 00:01:23,630 --> 00:01:28,210 Run claude ultrareview, or the ultrareview slash command inside a session, 25 00:01:28,370 --> 00:01:29,770 and answer the trust prompt. 26 00:01:30,270 --> 00:01:30,470 Done. 27 00:01:31,230 --> 00:01:33,030 The scripting side is where the detail is. 28 00:01:33,410 --> 00:01:37,070 You pass the trust workspace flag, which is spelled with two hyphens, 29 00:01:37,310 --> 00:01:38,550 trust, hyphen, workspace. 30 00:01:39,070 --> 00:01:40,530 And this flag takes no value. 31 00:01:41,050 --> 00:01:44,950 That matters, because ultrareview also takes a PR number or a base branch. 32 00:01:45,550 --> 00:01:50,950 Yeah, and the changelog is explicit about this: put the PR number or base branch 33 00:01:51,090 --> 00:01:52,270 BEFORE the flag. 34 00:01:52,810 --> 00:01:59,390 So you'd write claude ultrareview, then 1234, then the trust workspace flag. 35 00:01:59,970 --> 00:02:03,270 Or claude ultrareview develop, then the flag. 36 00:02:04,090 --> 00:02:08,750 If you want findings posted straight to the pull request, it's claude ultrareview 37 00:02:09,070 --> 00:02:13,370 1234, then the post flag, then the trust flag at the end. 38 00:02:14,030 --> 00:02:17,930 If you stick the flag first, you risk the parser treating your PR number as 39 00:02:17,970 --> 00:02:21,290 something it shouldn't, and that's a rubbish way to burn a CI run. 40 00:02:22,007 --> 00:02:23,167 There's also a second route. 41 00:02:23,427 --> 00:02:24,787 Set the environment variable 42 00:02:24,867 --> 00:02:31,747 CLAUDE_CODE_DISABLE_ULTRAREVIEW_TRUST_CHECK 43 00:02:31,807 --> 00:02:32,287 to 1. 44 00:02:32,907 --> 00:02:36,247 That's handy if you've got a dozen scripts and you don't want to edit every command 45 00:02:36,287 --> 00:02:36,527 line. 46 00:02:36,987 --> 00:02:38,527 Put it in your runner config once. 47 00:02:38,827 --> 00:02:40,687 Same effect, no per command flag. 48 00:02:41,342 --> 00:02:42,582 Now the caveats. 49 00:02:43,382 --> 00:02:46,602 First, that env var is a blunt tool. 50 00:02:47,122 --> 00:02:51,662 Setting it globally on a public repo workflow, especially one that runs on pull 51 00:02:51,702 --> 00:02:55,762 requests from forks, means you're switching off the check for code you don't 52 00:02:55,802 --> 00:02:56,422 control. 53 00:02:57,302 --> 00:03:02,822 The advice is to keep trust flags for internal branches and repos you've vetted. 54 00:03:03,862 --> 00:03:05,562 Second, billing and access. 55 00:03:06,282 --> 00:03:12,302 Ultrareview needs claude.ai authentication, so it's not available on Bedrock or 56 00:03:12,382 --> 00:03:17,642 Vertex setups, or with Zero Data Retention or HIPAA configurations. 57 00:03:18,642 --> 00:03:22,502 The brief says Pro and Max users get three free one time runs, 58 00:03:22,902 --> 00:03:27,662 then it uses usage credits, roughly five to twenty five dollars depending on diff 59 00:03:27,722 --> 00:03:28,062 size. 60 00:03:28,922 --> 00:03:31,582 Worth knowing before you wire it into a nightly job. 61 00:03:32,157 --> 00:03:32,637 Good point. 62 00:03:33,137 --> 00:03:35,137 Now a few smaller things from the same release. 63 00:03:35,817 --> 00:03:41,357 In bypass permissions mode in the Desktop app and VS Code, a dangerous rm or rmdir 64 00:03:41,477 --> 00:03:44,877 used to throw a prompt, and an unattended session would sit waiting. 65 00:03:45,577 --> 00:03:48,437 Now the command is denied straight away with a rewrite hint, 66 00:03:48,677 --> 00:03:50,537 so the agent can adjust and keep going. 67 00:03:51,217 --> 00:03:53,017 One to watch for on the sed side. 68 00:03:53,517 --> 00:03:59,357 Sed with the i flag, the in place edit, now gets the standard Bash prompt and runs 69 00:03:59,417 --> 00:04:03,537 exactly as written, instead of that simulated file edit preview. 70 00:04:04,157 --> 00:04:08,977 The catch is that, like other Bash edits, it's no longer restored by slash rewind. 71 00:04:09,397 --> 00:04:13,697 So if you lean on rewind as your safety net, commit before you let it loose. 72 00:04:14,263 --> 00:04:14,623 Two more. 73 00:04:15,103 --> 00:04:18,863 A nested CLAUDE.md over the size limit now loads only its start, 74 00:04:19,323 --> 00:04:23,123 which fixes sessions that kept compacting over and over after reading a huge one. 75 00:04:23,903 --> 00:04:28,563 And there's a security fix: on macOS and Linux, sandboxed Bash commands were running 76 00:04:28,583 --> 00:04:32,023 the shell's startup files outside the sandbox before each command. 77 00:04:32,563 --> 00:04:33,063 That's fixed. 78 00:04:33,715 --> 00:04:35,255 Last one's for VS Code folks. 79 00:04:35,715 --> 00:04:41,395 The wait for new problems after each Edit and Write in the chat panel is now 0.75 80 00:04:41,475 --> 00:04:45,455 seconds in most cases, down from a second to a second and a half. 81 00:04:46,255 --> 00:04:47,575 Small, but you'll feel it. 82 00:04:48,235 --> 00:04:52,935 So grep your pipelines for ultrareview today, and fix the flag order while you're in 83 00:04:52,975 --> 00:04:53,115 there.