Describe the Bug
Stopping a thinking agent during its thought process to supply key information midstream is completely ignored by the agent on the next message. The agent continues its original chain of thought without any regard for new messages inserted midstream.
Steps to Reproduce
In this particular case, using Claude 4 Sonnet MAX. I would prompt a thinking agent with information. Based on the agent’s visible thought process, I realized I forgot to include key information. So, I “Stop” the agent, and then enter that key information into the next message and send it.
Expected Behavior
Agent continues on from the new message supplied. Instead, agent seems completely unaware of any new message inserted after manually stopping. Asking the agent literally if it was aware of my “addendum” message says no.
Operating System
MacOS
Current Cursor Version (Menu → About Cursor → Copy)
Version: 1.1.3
VSCode Version: 1.96.2
Commit: 979ba33804ac150108481c14e0b5cb970bda3260
Date: 2025-06-15T06:35:49.230Z
Electron: 34.5.1
Chromium: 132.0.6834.210
Node.js: 20.19.0
V8: 13.2.152.41-electron.0
OS: Darwin arm64 24.5.0
Additional Information
This used to work really well - I was always able to stop an agent generating, insert a new message, and have it pick up from there instead. Now, it’s as if any messages inserted after a “Stop” command get completely ignored - I have to either wait for it to complete the original generation off the original message, or restart a new agent session and include the new details in the original message.