running-eval-driven-skill-development

What problem does it solve?

This Skill helps authors prove that a new Claude Code skill actually works before they ship it, instead of relying on a superficial review.

Core Features & Use Cases

• Evals-first workflow: Drafts three evaluation scenarios before the skill body so the behavior is defined up front.
• Checkable rubrics: Requires each scenario to use exactly three third-person, pass-or-fail assertions.
• Verification loop: Instructs the author to run a subagent against each scenario, score the response, and revise if thresholds are not met.
• Use case: Use it when writing a judgment-heavy skill such as an auditor, selector, diagnostician, or recommendation engine.

Quick Start

Use this skill to walk me through authoring a new non-trivial skill with three eval scenarios and a verification pass.