Editor's selection
GitHub Copilot Evals preview: repeatable tests for agent skills, MCP tools and coding workflows
A single successful agent response is weak evidence. Copilot Evals lets development teams repeat tasks and inspect what changed after customizing tools or skills.