This has only just started happening to day, but this subagent has asked me about 8 times for permission so far…
It’s never asked like this before and is formatted differently to other permission requests where it showed you the command.
Also, when it searches through a codebase it has never asked to do that before.
For AI issues: which model did you use?
Model name (e.g., Sonnet 4, Tab…)
For AI issues: add Request ID with privacy disabled
Request ID: f9a7046a-279b-47e5-ab48-6e8dc12daba1
For Background Agent issues, also post the ID: bc-…
Additional Information
Add any other context about the problem here.
Does this stop you from using Cursor?
Yes - Cursor is unusable
Sometimes - I can sometimes use Cursor
No - Cursor works, but with this issue
The more details you provide, the easier it is for us to reproduce and fix the issue. Thanks!
Hey @Ryan_9, thanks for the report, and sorry about the prompt spam.
This is a known issue we’re already tracking, and I’ve added your report to it. When Auto-Review is your run mode, a subagent (like the Explorer one in your screenshot) isn’t currently getting the same auto-approval treatment as the main chat, so each command it runs that isn’t on your allowlist stops and asks. That’s also why the prompt looks different: on the subagent card we summarize the command by the programs it calls (cd, rg, Get-ChildItem, …) instead of showing the full command line. You can expand the subagent to see the exact command before allowing it.
The fix is already in the next release, so this should go away once you update to 3.21 (due out soon)
Thanks @Ryan_9, and sorry you’re having to deal with this! I am sharing details or your agent requests with the team so we can dig in and figure this out.
Hi all! Can you all try to upgrade to 3.22.12 and letting us know if you’re still running into this issue? Apologies for the delay on getting this fix out!