vercel-optimize (Vercel): a Claude skill that audits Vercel cost and speed

Vercel's vercel-optimize skill has Claude audit a deployed project's cost and speed from real Vercel metrics and rank the fixes.

Installs: 118.2K; on skills.sh
Stars: 33.4K; on GitHub
Files: 60+; besides SKILL.md
Updated: Oct 2026; last repo commit

vercel-optimize runs a metrics-first audit of a deployed Vercel project. Claude collects your Vercel metrics, usage and project config, scans the code, and only then investigates the routes that the numbers point to. You get ranked recommendations with file references and version-aware doc citations.

You need the Vercel CLI v53 or newer, a logged-in session, a linked project folder and Node.js 20+. Route-level recommendations also need Observability Plus on the project. The skill bundles 155 files, mostly Node scripts, and runs them on your machine.

Good for

Teams on Next.js, SvelteKit or Nuxt who want a lower Vercel bill
Developers chasing slow or expensive routes and caching gaps
Engineers who want recommendations tied to metrics and verified files

Install vercel-optimize

Claude Code

Copy this prompt.

Install the vercel-optimize skill in Claude Code. Run this command:

npx skills add https://github.com/vercel-labs/agent-skills --skill vercel-optimize

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

From its listing on skills.sh. Source.

What's in the skill

When it loads

Use for Vercel cost and performance optimization on deployed projects, especially Next.js, SvelteKit, Nuxt, and limited Astro apps. Collect Vercel metrics, usage, project config, and code scan results first; investigate only metric-backed candidates; produce ranked recommendations grounded in verified files and version-aware Vercel/framework docs. Trigger for Vercel bill reduction, slow or expensive routes, caching opportunities, Function Invocations, Build Minutes, Fast Data Transfer, Core Web Vitals, Bot Management, Fluid compute, or cost breakdown requests.

Sections in SKILL.md

Prerequisites
Framework Support
Run Directory
Pipeline
1. Collect, scan, and merge signals
1.1 Stop on blockers
2. Gate candidates
2.1 Ask about audit scope when needed
2.2 Deep-dive and reconcile
2.3 Generate briefs and investigate
2.4 Collect outputs
3. Verify recommendations
4. Render report and final message
Recommendation Rules
Scanner Rules
Final Customer Terms

Key facts

Author: Vercel
Repo: vercel-labs/agent-skills
Skill folder: skills/vercel-optimize
skills.sh installs: 118,236
Repo stars: 33,354
License: MIT (stated in the skill's README)
Last repo commit: 2026-10-28
Bundled files: 155 files besides SKILL.md
skills.sh audits: Gen Agent Trust Hub: pass; Socket: pass; Snyk: pass

What it includes

collect-signals.mjs, scan-codebase.mjs and merge-signals.mjs build signals.json
Pulls 14 days of Vercel metrics, usage and contract data through the Vercel CLI and scans your code.
Runs code · 1. Collect, scan and merge signals
gate-investigations.mjs picks which routes and settings deserve a look
Default budget is 6 code-scope candidates, expandable with --max-candidates.
Runs code · 2. Gate candidates
Checks the metric evidence and drops disproven candidates
Runs code · 2.2 Deep-dive and reconcile
One brief per candidate; sub-agents when there are 3 or more
Read · 2.3 Briefs and investigation
verify-and-regen.mjs checks files, citations and version fit
Runs code · 3. Verify recommendations
render-report.mjs writes report.md and the final message
Runs code · 4. Render report

From the official docs

Permissions

Medium risk

Claude runs scripts and the Vercel CLI against your real account and reads your project code. Confirm the team and project it picked before it collects anything.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

It reads data from your Vercel account
It collects usage, metrics and contract data for the linked project. If it asks which project or team you mean, answer exactly; do not let it guess.
It runs many bundled scripts
The skill ships 155 files, mostly Node scripts. Read the commands Claude proposes before you approve them.
Linking a project changes local files
If the folder is not linked, SKILL.md has Claude run vercel link, which writes a link file in your project. Check the project and team name first.

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details

Turning on read-only	No read-only switch is named in SKILL.md. Its output is a report; the README says reports can include before/after code.
Talks to	Vercel API through the CLI. Your AI provider receives the metrics and code snippets Claude reads
Sign-in and keys	No account or key is named in SKILL.md.
Scopes	Runs the Vercel CLI (metrics, usage, contract, api) in your linked project Runs the bundled Node scripts and writes results to a temporary run folder Reads your project code and package.json

Questions

What does vercel-optimize do?

It audits a deployed Vercel project for cost and performance. It starts from real production metrics, investigates only metric-backed candidates, and returns ranked recommendations grounded in verified files and Vercel or framework docs.

How do I install vercel-optimize in Claude Code, Claude or Codex?

Use the install box at the top of the page and pick Claude Code, Claude or Codex. Claude Code and Codex read skills from a folder on your machine. In the Claude app you upload the skill folder as a zip.

What do I need before I run it?

Vercel CLI v53 or newer, a vercel login session, a project folder linked with vercel link, and Node.js 20+. Route-level recommendations need Observability Plus on the project.

What if I do not have Observability Plus?

Claude stops and gives you two choices: enable Observability Plus and rerun, or accept a limited code-only audit. SKILL.md says it must not fall back to code-only silently.

Which frameworks does it support?

Next.js (App and Pages Router), SvelteKit and Nuxt are supported. Astro is limited. Hono, Remix and unknown frameworks are blocked by default unless you accept a limited platform and code audit.

Is it safe to use?

It runs Node scripts and Vercel CLI commands on your machine and reads your project and Vercel account data. SKILL.md says never to put auth tokens in shell commands. Check which team and project it is about to audit.

Information

updated Nov 2, 2026

Repository: vercel-labs/agent-skills
Folder: skills/vercel-optimize

vercel-labs/agent-skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 3

vercel-optimize SKILL.md · github.com
vercel-optimize on skills.sh · skills.sh
vercel-optimize README · github.com