If you’re building agentic workflows, prompts, skills or custom agents and they’re tied to a specific model, now is a good time to add evaluations to the stack.

You can do it manually every now and then, but at some point you want an evaluation harness running continuously.