This scaffolds eval.yaml test files for agent skills in the dotnet/skills repository, handling everything from fixture organization to rubric definitions. You'd reach for it when creating new skill tests or adding evaluation scenarios, not for running or debugging existing ones. The workflow is thorough: it enforces directory conventions, guides you through assertions versus rubric items, and includes an overfitting classifier that judges whether your rubric tests outcomes (good) versus techniques or vocabulary (bad). The distinction matters because rubric quality dominates the composite improvement score at 40% weight. It's opinionated about test structure but for good reason, especially the guidance on neutral prompts and broad assertions that don't gate on specific syntax.
npx -y skills add dotnet/skills --skill create-skill-test --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
wshobson/agents
github/awesome-copilot