Which method will be most effective in controlling Cursor Agent's unwanted changes?
4
190Ṁ92
Apr 20
7%
Using a .cursorrules file
37%
Configuring Yolo mode guardrails
10%
Switching to a different AI model
20%
Using smaller, more focused prompts
7%
Enabling Privacy Mode
10%
None of these methods were effective
10%Other

This is all AI generated. I found Cursor to make too many unwanted changes... I added a .cursorrules file but so far it didn't seem to have done the trick. I'm looking for help, convince me!

This market asks which method will prove most effective in controlling Cursor Agent's unwanted changes based on my experience by July 31, 2025.

I'll resolve this market based on which method I find most effective in my personal usage. If I haven't found any effective method by the resolution date, this will resolve to 'None of these methods were effective'.

Sources:

Get
Ṁ1,000
to start trading!
Sort by:
boughtṀ10 YES

@ikoukas thanks, I was hesitant about Yolo mode. With these in place, that barrier gets lower:

@ozan glad to hear it!

I'm getting good results with full Issue descriptions from GitHub. I let ChatGPT write the issues based on short prompts, not longer than a few sentences/prompts.

I'll vote a little bit against the "using smaller" to express it, but not enough to sway it to my advantage.

The market by nature is vague in terms of resolution criteria, but I'm open to figuring out a few test cases so that we can settle together.

This is large to me: "Let's add a debug mode as a hook to this feature"

Cursor doesn't even seem to know how to create rules for itself?
https://x.com/ozanogred_en/status/1904092571833442390

sold Ṁ10 YES

Which ai model do you use?

@ian claude-3.7-sonnet was the default, so I keep using that.

I’ve found it ignores my cursorrules files too!

@ian I've seen a few times when the agent pasted the contents of the file in it's process and made use of it.

© Manifold Markets, Inc.Terms + Mana-only TermsPrivacyRules