benchmark-e2e

Was der Skill vermittelt

This skill drives an end-to-end benchmark of the vercel-plugin from inside its own repository. A single Bun script creates sample projects, runs them through Claude Code in print mode so the plugin can inject skills, starts their development servers to check that they work, analyses the conversation logs and writes an improvement report. It describes the run manifest, event and report files it produces, a quick mode over three projects, and an overnight loop meant to improve the plugin over time. It is tooling for the plugin's maintainers.

Funktionen

Details