What you'll be able to do
- Write an agent brief that gets a usable result on the first try
- Run two or three agents in parallel without them colliding
- Use
/best-of-nto compare several attempts and keep the best one - Know when to move work to the cloud, or hand a large effort to Cursor Projects