Vol III · Section 2
Agents in production
Reliability, cost, evaluation, security and human oversight for agents that run for real.
1How do you evaluate an agent? Outcome vs trajectory evaluation.2How do you make agent tool calls reliable — retries, timeouts and idempotency?3What is durable execution, and why do long-running agents need it?4How do you control agent cost and token budgets?5How do you reduce latency in an agent?6How do you defend an agent against prompt injection? Where does least privilege come in?7How should an agent execute code safely? Talk about sandboxing.8How do you design human-in-the-loop approvals for an agent?9What does observability look like for agents?10How would you roll out a new agent to production safely?