Cursor started plan implementation and editing files without leaving Plan mode

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

I was in Plan mode and without pressing the Build button Cursor started to build but leaving the plan unbuilt since the TODOs are still unchecked and the Build button is still there but the code has been produced

Steps to Reproduce

I was planning with Opus 4.5 and then reviewing the plan with Codex 5.2 XHigh. The model asked me if I wanted to include its findings in the plan and I said “Do it” and instead of changing the plan it started to build.

Operating System

MacOS

Current Cursor Version (Menu → About Cursor → Copy)

Version: 2.3.41
VSCode Version: 1.105.1
Commit: 2ca326e0d1ce10956aea33d54c0e2d8c13c58a30
Date: 2026-01-16T19:14:00.150Z
Electron: 37.7.0
Chromium: 138.0.7204.251
Node.js: 22.20.0
V8: 13.8.258.32-electron.0
OS: Darwin arm64 25.2.0

For AI issues: add Request ID with privacy disabled

Request ID: 4c6ddc88-2113-4a58-801b-ff3b168a61d0

Does this stop you from using Cursor

No - Cursor works, but with this issue

Hey, thanks for the report.

This looks like a command interpretation issue in Plan mode. The model took “Do it” as a signal to start implementing, instead of modifying the plan. This is a known class of bugs with auto-switching between modes, and we’ve had a few similar reports.

Could you share:

  • A screenshot of the Plans UI state (TODOs unchecked and the Build button still there)
  • The full chat transcript from when the model suggested “include findings” up to when the implementation started

That’ll help the team reproduce and fix the issue more accurately.

Tried to find it but I’ve created many chats since then and there’s no way to search by chat content

1 Like

I understand - finding an old chat without search is really a problem.

Your report is still useful though. Even without the full transcript, the fact that “Do it” triggered implementation instead of changing the plan - that’s a specific pattern that helps the team see where the model misinterprets commands.

For future bugs, a couple of tips:

  • Take a screenshot right when you see the problem (especially UI state - checkboxes, buttons)
  • You can copy the request ID straight from the chat context menu before switching

Thanks anyway, I really appreciate it

This topic was automatically closed 22 days after the last reply. New replies are no longer allowed.