Any way to let Claude run /verify and /code-review on its own again?
First off, thanks for these two — I lean on them a lot.
I noticed after updating to 2.1.215 that Claude won't run /verify or /code-review by itself anymore, so I've been typing them in by hand every time. I'm guessing that's the better default for most people, and honestly the automatic reviews were probably noise for a lot of folks.
My problem is that I'd built my whole routine around Claude doing it without me. It makes a change, checks its own work with /verify, looks over the diff with /code-review, fixes whatever comes up, and only then commits. Now that middle part just stops and waits for me. It's a small thing each time, but it adds up, and it turned something that used to run on its own into something I have to sit and shepherd.
I went looking for a setting to turn it back on just for myself and couldn't find one. As far as I can tell the two skills don't even show up as options for Claude now, so asking it to "just run /verify" partway through a task doesn't work either.
I know I'm probably an edge case and I'm not trying to argue with the default. I just wanted to ask if there's an option I've overlooked, and if not, whether letting people opt back in would be something you'd consider. If the answer's no, no hard feelings — figured it was worth asking.
Thanks for all the work on this.
Environment: 2.1.215, native install, Linux.
Showing cached comments. Read the full discussion on GitHub ↗
6 Comments
Disabling model invocation for code-review is disappointing. Can we revert this change?
This was a random, unnecessary, breaking change for people using this skill in automated workflows. I am already implementing a workaround for this but +1, bad idea.
I have also found this to be an unhelpful regression. I understand maybe changing the defaults, but it needs to be possible to configure this, and ideally I need a way to persist any setting so that all of my local and cloud sessions are able to auto-run code-review.
I have a Claude hook which prompts my agents to review a checklist every time it pushes a git branch or creates a PR. I have worked around this issue by changing this checklist:
Before:
---
/code-review highand address the findings before the push/PR. The trigger is completeness: if you already know more edits are coming this session, skip it now and run it at the next push/PR rather than reviewing a half-finished diff.---
After:
---
higheffort by following the instructions in claude-code's code-review plugin, and address the findings before the push/PR. Read that file and follow it rather than invoking the/code-reviewskill: claude-code#79282 currently blocks Claude from running the skill directly. The trigger is completeness: if you already know more edits are coming this session, skip it now and run it at the next push/PR rather than reviewing a half-finished diff.---
This is an extremely counterproductive regression, our team was using this as part of a larger review process, gathering results from this and internal tasks.. now it's all broken randomly with no warning.
Reviewing just became a lot more annoying.