Prompt Evaluation

This skill is a router and workflow. It teaches eval-driven prompt refinement, then dispatches to one reference file for the chosen grading approach or tool. Read the references on demand — do not pre-load all of them.

When this matters

Prompt engineering without evals is guesswork. Two prompts can both "look good" on a handful of cherry-picked inputs and still differ by 20+ points on a real test set. The whole point of this skill is to replace "v2 feels better than v1" with a number you can defend.

The user wants one of these: