I adjust some of my custom rules, to make it more aligned for what I want to do and now it’s super effective, I just guide it and I monitor it in every interaction to avoid problem in the future, my custom rules is not perfect and it’s sometimes not following the rules but if you make some adjustments to the rules, you can make your rules super effective.
Yeah, because the rules i customize is to not go to the agent mode when the confidence level is not 95%+ because it needs more clarification to ensure the act mode you need to answer all of the question and read the flow
You can adjust it to 85 percent tho, you have my custom rules you can modify it freely
I actually replace it to agent instead of act mode accidentally but you can change it anytime, and if it’s not following your rules you need to make sure you are going to put a context window all of the rules and md files that I have on my github repo
Yes, you need to use composer agent chat to use the full potential of this rules
It’s more consistent when using both rules to make sure the AI are doing it the exact instruction rules, that’s why it’s more powerful with these new project rules are combined, i honestly want the cursor to add a rules and it looks like this " plan.cursorrules " same function by the .cursorrules, it would be awesome and powerful if they add this feature
About to start a new project, going to give this a shot and I’ll report back. If anyone has used cursor extensively, this is addressing a pretty big headache.
I love the idea of creating a memory file and having keywords.
Me: “Where is this variable defined?”
Cursor: “I see the problem. Let me create 4 new classes and refactor the entire backend”
I have a new memories format this is more less lines that will create by AI so it will be going to be more efficient, and memories will not be filled up immediately