Is it Cursor or Sonnet 4 feature for almost all requests, it’s like oh well let me test this out and it will write a test script and one it’s done testing it will delete that script. Also it unecessarily keeps on starting my server and always with incorrect command, like it uses standard for everything and if my start command is not standard it won’t work and will keep on iterating. Like I use yarn but it always starts the server with npm etc. Also will try to test things which are not testable, it’s overly cautious and tries to test out things. Is is happening with everybody?
Hi, no, its not happening to everybody 
Usually AI (Claude 4 Sonnet) tries what it found in training data (e.g. people using node use npm for example)
Did you create project rules and describe your requirements?
- What it must do
- How it should do things
- What it should skip
Unless you describe what is required it wont know what you need. So it takes safe assumptions and makes tests or starts servers,…
Maybe you can share more what you are asking AI to do and how you ask it
1 Like
Yes, I do face same problem. I have no cursor rules or anything. But it keep writing me a test_file for both NodeJS and FastAPI backend lol
And sometimes it just print in console log "Test 1 : Successfully Executed " without proper testing functions lol
no cursorl rules
`=== Testing Agent Return Type Fix ===
Dict input: {‘output’: ‘Current date and time: 2025-05-26 08:00:00 UTC’, ‘intermediate_steps’: }
Extracted output: Current date and time: 2025-05-26 08:00:00 UTC
Output type: <class ‘str’>
String input: Direct string response
Extracted output: Direct string response
Output type: <class ‘str’>
Current time tool result: Current date and time: 2025-05-26 08:05:29 UTC
Time result type: <class ‘str’>
Fix verified: All outputs are strings
`