The agent sends keyboard interrupts to the terminal randomly while a command it generated is running. It then derails itself trying to debug the errors generated
Steps to Reproduce
Not reproducible. Random occurences every ~20 prompts
Expected Behavior
It allows the command to complete, before thinking again, or responding
I have a memory explicitly saying βThe user prefers that the assistant not interrupt long-running commands and wait longer for them to complete.β
Is no one else running into this issue? I tried emailing Cursor and was only able to get their AI bot to reply. This is progressively getting worse and makes the agent unusable.
Update: this is still an issue even with the latest updates
Version: 1.3.5 (user setup)
VSCode Version: 1.99.3
Commit: 9f33c2e793460d00cf95c06d957e1d1b8135fad0
Date: 2025-07-30T00:37:52.749Z
Electron: 34.5.1
Chromium: 132.0.6834.210
Node.js: 20.19.0
V8: 13.2.152.41-electron.0
OS: Windows_NT x64 10.0.26100
Also happened to me.
it runs a command, then KeyboardInterrupt itself, then try to debug by changing files, and repeat, rapidly consuming the tokens without any progress.