vercel-optimize

Run this audit from your linked Vercel project directory. It collects usage and route metrics first, then ranks recommendations such as fixing low cache hit rates, narrowing middleware matchers, enabling Fluid Compute, or reducing ISR over-revalidation, each tied to observed data.

The skill queries route-level function duration, invocations, and cold starts through the Vercel CLI, then applies deterministic thresholds to flag hot routes. Only routes that pass these metric gates are investigated in source code.

Metric-backed route-to-file recommendations are strongest for Next.js App and Pages Router and SvelteKit, supported with generic checks for Nuxt, and limited for Astro. Hono, Remix, and unknown frameworks pause by default unless you accept a limited platform-only audit.

You need Node.js 20+, Vercel CLI v53 or later, an authenticated CLI session, and a project directory linked with vercel link. Route-level metric analysis also requires Observability Plus; without it the skill pauses and offers a limited code-only audit.

The skill is observability-first: without route metrics it cannot rank hot routes or prove cost impact, so it pauses instead of guessing. You can enable Observability Plus and rerun, or explicitly accept a limited scanner-only audit.

No. Cost impact is framed as order-of-magnitude phrases rather than precise dollar figures, because exact savings cannot be verified from the available signals. Recommendations always cite observed metrics and allow-listed documentation instead.