Claude Code plugins can now be tested for better answers

Claude Code now lets developers test whether a plugin or skill actually improves Claude’s answers, instead of relying on subjective prompt checks. That matters as the ecosystem grows beyond isolated experiments and into larger collections of reusable workflows.

The workflow starts with claude plugin eval init, which asks what a good result looks like and writes test cases in the plugin folder. Running claude plugin eval . then evaluates the extension against those cases.