benchmark-testing

To benchmark a plugin with Claude Code, this Skill automates creating isolated test directories, installs the vercel-plugin in each, and spawns WezTerm panes running Claude Code sessions with crafted prompts to validate skill injection behavior deterministically.

Automated skill injection testing works by applying a workflow across multiple isolated directories, installing the vercel-plugin in each project, and launching Claude Code sessions with crafted prompts to exercise and observe plugin activation in controlled environments.

Yes, you can run benchmark tests for the Vercel plugin without external inputs. The workflow operates entirely within the local environment, requiring no external dependencies beyond the local setup to ensure deterministic, isolated, and reproducible test results.

The best way to validate Vercel plugin wiring and activation is by spawning Claude Code sessions in WezTerm panes with crafted prompts. This exercises skill injection in isolated test directories, allowing you to observe plugin behavior in controlled, reproducible environments.

Yes, WezTerm is required to spawn the Claude Code sessions for benchmark testing. The workflow uses WezTerm panes to launch the sessions with crafted prompts that exercise the vercel-plugin skill injection in each isolated test project.

Inconsistent benchmark test results often stem from poor isolation or non-deterministic setup. This Skill ensures deterministic setup and reproducible results by creating isolated test directories for each benchmark run, preventing cross-contamination between different plugin configurations.