Cost Budget

Why cap each agent run with a cost-per-task budget?

How do you enforce a $0.50 per-task LLM budget at runtime beyond setting max_tokens?

Name the three canonical agent stopping conditions and explain why cost cap is the hardest to implement correctly