Skip to content
Kitbay
Submit

benchmark-agents

Apache-2.0Updated on Nov 10, 2026333GitHub stars

A maintainer skill in the vercel-plugin repository for testing how the plugin behaves during advanced agent builds.

What the skill teaches

A maintainer skill in the vercel-plugin repository for testing how the plugin behaves during advanced agent builds. The agent creates test folders, installs the plugin, opens several Claude Code sessions in WezTerm panes with prompts that exercise Vercel features like AI Gateway, Queues, Sandbox, Flags, the Chat and Workflow SDKs and MCP, and then reads debug logs to see which skills the plugin injected. It finishes with a coverage report and a loop of fixes and releases. It is meant for people working on the plugin, not for app development.

Capabilities

  • Launch Claude Code test sessions with the plugin installed
  • Check which plugin skills were injected in each session
  • Write a skill coverage report for advanced scenarios