The original title is "How to Know If Your Claude SKILL.md Actually Works"

Short summary

The author introduces skilleval, a CLI tool for objectively evaluating Claude SKILL.md files using blind A/B testing and a randomized LLM judge. It features deterministic assertions, multi-turn conversation support, regression detection, and CI integration to prevent silent skill degradation. The open-source tool aims to replace subjective 'vibes-based' skill evaluation with repeatable, margin-based scoring.

• •Introduces skilleval, a CLI for objective Claude SKILL.md evaluation
• •Uses blind A/B testing, randomized LLM judge, and deterministic assertions
• •Supports regression detection, version comparison, and CI integration

Generated with AI, which can make mistakes.