Spins up a focused sub-agent tuned for your specific task—research, writing, analysis—with automatic model selection and built-in self-critique to catch errors before returning results.
Best for: When a single prompt isn't enough and you need a mini-expert to own a multi-step problem.