Experience designing agent evaluation paradigms, including trajectory evaluations, LLM-as-judge workflows, task-success metrics, tool-call correctness checks, rubric-based qualitative grading, adversarial scenario testing, regression eval suites, and human-in-the-loop review. Design tooling that supports modern agent patterns, including structured tool calling, filesystem-using agents, memory and retrieval, planning loops, subagents, agent graphs, and human-in-the-loop workflows.