Let Your Agent Cook: Using Skills to Evaluate and Improve Your App — Ankur Duggal, Arize AI

Why watch

Ankur Duggal demonstrates how to use Arize AI's CLI and skills to connect a financial agent's failures (hallucinations, redundant tool calls, wrong summaries) to a coding agent for investigation. The demo shows tracing through supervisor and specialist agents, pulling evaluation results into a local workspace, identifying grounding problems and fabricated responses, and using dataset/evaluator/experiment skills to test proposed fixes. Ideal for developers building AI agents who need better debugging and evaluation workflows.