What does your
agent need
to get right?

Describe the workflow and the result you need. Scope an environment, an evaluation, or an environment with RL training and retesting.