eval-driven-development

To measure prompt quality improvements, this Skill helps you define an evaluation rubric, score the baseline prompt, implement changes, and compare the before-and-after outcomes quantitatively. It provides a structured workflow to verify that modifications actually enhance performance against your defined criteria.

Rubric-driven evaluation is a workflow where you define explicit quality criteria before making changes to artifacts like skills, prompts, or configurations. It establishes a baseline measurement and then compares post-change outcomes to ensure improvements are measurable and aligned with your standards.

You can compare baseline and post-change outcomes by first scoring the current configuration against a designed rubric. After implementing the configuration updates, the Skill evaluates the new version and contrasts the scores to validate whether the changes produced the desired quality gains.

Yes, you can use this workflow to evaluate Claude Code configurations and workflows. It is applicable to any artifact where explicit before-and-after evaluation is required, allowing you to validate quality assurance improvements against predefined rubric standards.

The best way to establish a baseline is to define your evaluation rubric first, then apply it to score the current prompt before making any modifications. This baseline score serves as the quantitative reference point for comparing post-change results.