Hey folks,
I am a professional trainer, and I’d be eager to run experiments on a regular basis to demonstrate the effects of using Cursor features. For instance, showing students how indexing a documentation will improve the results for a given prompt, how I get better results with a skill and so on.
The problem is that it’s difficult to assess right now because I am never really in a clean state : I can’t tell if this or that doc was indexed, if the agent did read some code etc.
There are features that may help but they are scattered, like I could add/remove an indexed doc, test multiple models on a prompt etc. but that’s still not super clean.
Is there an easy way for instance to create independent profiles, that will have a totally fresh config and context?
# Default Cursor data path is %APPDATA%\Cursor
# Create experiment-specific directories:
# "Baseline" – no docs indexed
cursor --user-data-dir="$env:USERPROFILE\.cursor-baseline"
# "With Docs" – add docs in this instance only
cursor --user-data-dir="$env:USERPROFILE\.cursor-with-docs"
# "With Skill" – preconfigure skills in this instance
cursor --user-data-dir="$env:USERPROFILE\.cursor-with-skill"
# "Fresh" – always empty for a clean run
cursor --user-data-dir="$env:USERPROFILE\.cursor-fresh-$(Get-Date -Format 'yyyyMMdd-HHmm')"
But it is also possible to copy the files, and then restore from backup to start fresh. Normal path in Windows is $env:APPDATA\Cursor\User
Did you try going to Cursor Settings → Indexing & Docs tab? You can sync and delete your index from there.
Also if you want to demonstrate with fresh state - you should create a new folder on your computer and open cursor inside that. Then place your project files into that.
