benchmark-testing

To benchmark plugin injection, create a base directory, generate subdirectories for each test, install the plugin in every directory, and launch Claude Code instances with realistic prompts to validate behavior under varied conditions.

Vercel-plugin skill injection testing validates how plugins behave when injected into Claude Code sessions using realistic, technology-agnostic prompts. It exercises plugin behavior across isolated directories to ensure correct execution under varied prompt conditions.

WezTerm is required to launch multiple Claude Code sessions in separate panes for benchmark testing. This setup allows simultaneous execution of realistic prompts across isolated directories to validate plugin injection behavior.

Benchmark testing uses realistic, technology-agnostic prompts to exercise vercel-plugin skill injection. This approach ensures the validation process remains broadly applicable across different technology stacks during CI-like experimentation.

Running benchmark tests in isolated directories prevents cross-contamination between plugin injection tests. Each directory validates plugin behavior independently, ensuring accurate results during CI-like experimentation with varied prompts.