Content
@
0 reply
0 recast
2 reactions
Thomas Humphreys
@so
Cursor has been a game-changer for me. I let Composer agents run in the background to handle mundane tasks while I focus on more important/creative work.
1 reply
0 recast
5 reactions
Samuel ツ
@samuellhuber.eth
what kinds of tasks do you give them? I found them to be buggy at large codebase refactors
1 reply
0 recast
1 reaction
Thomas Humphreys
@so
I use them for pretty much anything. Just now, I had them set up a full-stack invitation registration flow for new passkey users. The server-side stuff went smoothly (the right prompt makes a huge difference). That said, I've noticed long conversational threads can start to degrade, probably because the Cursor team does some behind-the-scenes optimizations to cut costs. Right now, I'm experimenting with cross-repo agent context, trying to get my all repos within the org to communicate with each other. This should be a clear roadmap item for the team.
1 reply
0 recast
1 reaction
Samuel ツ
@samuellhuber.eth
How do you structure your prompts? What’s the usual task where you‘re line okay let’s run cursor agent
1 reply
0 recast
1 reaction
Thomas Humphreys
@so
Great question! Composer really excels when you have clear, end-to-end requirements for your repo. Start by setting a global goal, then break it down into specific needs, like focusing on security or other priorities. The key is making sure you reference all the necessary pages, docs, and anything else you want it to handle. If you do that correctly, it will auto fix any linting issues. Also, Cursor rules (https://dotcursorrules.com/rules) are great for tackling those pesky lint failures that keep popping up.
0 reply
0 recast
0 reaction
