Agentic Development · Evaluation
A test suite that runs an agent against a fixed set of tasks and scores its outputs automatically, so changes to a prompt or model can be compared side by side.
Example: Before shipping a prompt change, the team runs it against 50 saved support tickets and checks the pass rate didn't drop.
In practice: Teams lean on this before shipping a prompt change, since a fix for one task can quietly break three others without a way to check them all at once.