What problem does it solve? Vercel bills and slow routes are hard to diagnose because cost and latency signals live in observability data while fixes live in source code. This Skill connects the two by collecting Vercel metrics first, then investigating only the routes and files those metrics implicate, so every recommendation is backed by observed evidence rather than guesswork. ## Core Features & Use Cases - Metric-driven audits: Collects route-level function invocations, duration, cache hit rates, Core Web Vitals, ISR activity, and billing usage via the Vercel CLI before reading any source files. - Deterministic gating: Pure-JavaScript threshold gates decide which candidates deserve investigation, with a default budget of 6 code-scope candidates and an optional user-approved expansion. - Verified recommendations: Every recommendation is checked against real files, an allow-listed citation library, and the detected framework version before appearing in the final Markdown report. - Use Case: A team notices their Vercel bill doubled. They run the audit from the linked project directory, and the Skill identifies an uncached high-traffic route, an over-aggressive ISR revalidate interval, and a broad middleware matcher, each with before/after code and documentation citations. ## Quick Start From a linked Vercel project directory, ask your coding agent to optimize this Vercel project and reduce its cost and slow routes.
