benchmark-sandbox

Automating web application performance testing in Vercel Sandboxes involves provisioning ephemeral Firecracker microVMs to execute benchmark prompts and extract artifacts. This Skill orchestrates that pipeline, producing structured JSON scores and coverage reports for each eval phase without manual setup.

Vercel Sandboxes provide ephemeral Firecracker microVMs used to run isolated benchmarking scenarios. They offer a controlled environment for executing benchmark prompts, extracting hook artifacts, and generating coverage reports for performance analysis.

Run benchmark eval scenarios with Claude Code and Vercel CLI by executing the default script via your terminal. The pipeline automates building, verifying, and deploying apps within microVMs, using dynamic scenarios from JSON files and CLI flags for customization.

Yes, you need a Vercel account and access to Claude Code to run microVM benchmarking scenarios. These dependencies are required to provision the ephemeral Sandboxes, execute the eval pipeline, and extract the resulting artifacts.

Yes, you can customize benchmark prompts with CLI flags in Vercel Sandboxes. The execution script supports multiple CLI flags for customization and loads dynamic scenarios from JSON files to tailor the performance testing pipeline.

The 3-phase eval pipeline for performance testing consists of building, verifying, and deploying applications within ephemeral microVMs. It executes benchmark prompts across these phases to produce structured JSON scores and generate comprehensive coverage reports.