Kipt

Vercel optimize skill: what it does and how to install it

The Vercel optimize skill audits a deployed Vercel project's cost and performance from real metrics, then ranks verified fixes. What it does and how to install.

By the Kipt team · Updated 24 October 2026

Source: vercel-labs/agent-skills · Maintainer: Vercel · License: MIT (per the skill README and repo README; no LICENSE file in the repo) · Checked: 24 October 2026

What is the Vercel optimize skill?

The Vercel optimize skill has your coding agent audit a deployed Vercel project for cost and performance, starting from real production metrics rather than a read of the code. It finds slow or expensive routes, caching gaps, function and build-minute waste, then ranks fixes with file references and doc citations. It is for teams running Next.js, SvelteKit or Nuxt on Vercel who want to cut the bill or speed up hot routes.

It lives in the vercel-optimize folder of vercel-labs/agent-skills, and it is the first skill listed in the repo README. The description's first sentence reads: "Use for Vercel cost and performance optimization on deployed projects, especially Next.js, SvelteKit, Nuxt, and limited Astro apps."

As of 24 October 2026, its skills.sh page listed about 109,100 installs (snapshot, 24 October 2026). The skill is at version 1.2.0 and was last changed in June 2026.

What does it do?

The skill is model-invoked. Its description names triggers such as Vercel bill reduction, slow or expensive routes, caching, Function Invocations, Build Minutes, Fast Data Transfer, Core Web Vitals, Bot Management, Fluid compute and cost breakdowns. The README's suggested prompt is simply "optimize this Vercel project".

Its doctrine fits in four rules from the SKILL.md: metrics first, deterministic gates, candidate-bound scope and version-aware citations. The opening line is blunt: "Do not inspect source files until signals.json exists and a deterministic gate points to a route, file, or project setting."

The pipeline, all driven by Node.js scripts in the folder:

  1. Collect. collect-signals.mjs pulls metrics, usage and project config through the Vercel CLI over a shared 14-day window. scan-codebase.mjs runs code scanners and merge-signals.mjs joins the two into signals.json.
  2. Stop on blockers. Unsupported framework, unclear team scope, no traffic, missing Observability Plus or an unlinked folder each triggers a scripted stop: the agent asks you how to proceed, or links the folder and reruns collection.
  3. Gate. gate-investigations.mjs picks candidates with fixed thresholds. The default budget is 6 code-scope candidates; --max-candidates 12 or all widens it.
  4. Investigate. One brief per candidate. With 3 or more briefs the agent spawns one sub-agent per brief when the host supports it, and each reads only the files its brief names.
  5. Verify. verify-and-regen.mjs checks files, citations and version fit, withholds weak recommendations and plans retries.
  6. Report. render-report.mjs writes a Markdown report and a short final message, which the agent prints verbatim.

What's in the folder: SKILL.md, README.md, AGENTS.md (a cross-agent entry point), CONTRIBUTING.md, metadata.json, 15 scripts in scripts/, a lib/ of gates, scanners and sanitizers, and references/ with the doctrine, playbooks for content sites, ecommerce, SaaS, AI apps, API services, marketing and SvelteKit, support topics, and docs-library.json, the allow-list of citable docs.

How do you install it?

The skills.sh page gives this command. The skills CLI matches --skill against the frontmatter name, which here matches the folder:

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

The README writes the same thing in short form, npx skills add vercel-labs/agent-skills --skill vercel-optimize. Add -g for all projects or -a claude-code to target Claude Code. A global install lands in ~/.claude/skills/vercel-optimize/, one of the standard Claude Code skill locations.

Manual install, from the skill's README and AGENTS.md: copy skills/vercel-optimize into .agents/skills/vercel-optimize, then add a line to your project AGENTS.md telling the agent to follow that SKILL.md and collect Vercel metrics before reading source files.

Codex and Cursor read the same SKILL.md format; see Codex skills and Cursor skills. The repo ships no claude.ai ZIP for this skill, and it needs a local shell with the Vercel CLI, so a terminal agent is the fit.

How do you use it?

Run your agent from the linked app directory. Example prompts (suggestions, except the first, which is the README's):

  • optimize this Vercel project
  • Our Vercel bill jumped this month. Find out which routes and services drove it.
  • Which routes have the worst cache hit rate, and what should we change?
  • Check whether we are over-revalidating ISR pages.
  • Audit Build Minutes in this monorepo.

Requirements, from the SKILL.md and README:

  • Node.js 20 or later.
  • Vercel CLI v53 or later with vercel metrics, vercel usage, vercel contract and vercel api (npm i -g vercel@latest).
  • A logged-in CLI (vercel login) and a linked directory (vercel link). VERCEL_PROJECT_ID alone is not enough for vercel metrics.
  • Observability Plus for route-level, metric-backed recommendations.

The README adds a self-check: "If it starts by reading source files or guessing from vercel.json, the skill was not loaded correctly."

Which frameworks does it support?

Framework Status in the SKILL.md
Next.js App Router Supported, strongest route mapping
Next.js Pages Router Supported
SvelteKit Supported
Nuxt Supported, fewer framework-specific recommendations
Astro Limited
Hono, Remix, unknown Blocked by default; limited audit only if you accept

The README's roadmap lists Hono and Remix route mapping, AI Gateway, Sandbox, and Blob, Edge Config, Runtime Cache, Workflows, Queues, Flags and Microfrontends billing as planned.

Tips and limits

From the skill and Vercel's docs, not from testing:

  • Observability Plus is the real dependency. The vercel metrics docs say metrics other than Web Analytics and Speed Insights require it. Observability Plus is available on Pro and Enterprise, is on by default for teams created or upgraded to Pro on or after April 3, 2026, and costs $1.20 per 1 million events as of October 2026.
  • It needs traffic. With no meaningful traffic in 14 days, it says route-level metrics are sparse and limits itself to traffic-independent findings.
  • Scope must be unambiguous. If it can't tell which project and team to audit, it stops and asks rather than guessing from vercel whoami.
  • No dollar figures. The rules require "cost magnitude phrases only; never customer-facing $N savings."
  • Scripts run on your machine. The agent runs the bundled Node.js scripts, which call the Vercel CLI and read your repo, writing results to a temporary run directory. The skill forbids putting tokens in commands.
  • Audits. On 24 October 2026 the skills.sh page showed Pass from Gen Agent Trust Hub, Socket and Snyk.

The MIT License lets you copy and adapt the skill as long as the notice stays with it.

Save it in your Kipt library

In Kipt, open Skills and paste the skill's GitHub folder link into the import box. Kipt reads the SKILL.md and the folder's scripts and references; for a repo with many skills, like this one, it shows a list to pick from. You can also drop the folder or a .zip. Your skills live in your own private GitHub repo and install into Claude Code or Codex with one command:

npx skills add https://github.com/<you>/<repo>.git --all -g

Kipt is not affiliated with Vercel.

Similar skills

Questions

What is the Vercel optimize skill?

vercel-optimize is an official Agent Skill in Vercel's vercel-labs/agent-skills repo. It runs a cost and performance audit of a deployed Vercel project: it collects Vercel metrics, usage and project config first, investigates only the routes and files those metrics point to, and produces a ranked report of verified recommendations.

How do I install the vercel-optimize skill?

Run npx skills add https://github.com/vercel-labs/agent-skills --skill vercel-optimize, the command on its skills.sh page. The README also documents a manual install: copy the skills/vercel-optimize folder into .agents/skills/ and reference its SKILL.md from your project's AGENTS.md.

Does the Vercel optimize skill need Observability Plus?

For its main, metric-backed route ranking, yes. Vercel's docs say vercel metrics needs Observability Plus for metrics other than Web Analytics and Speed Insights. Without it, the skill stops and offers a limited scanner-only audit instead of silently falling back.

Which frameworks does vercel-optimize support?

The SKILL.md lists Next.js App Router, Next.js Pages Router, SvelteKit and Nuxt as supported and Astro as limited. Hono, Remix and unknown frameworks are blocked by default, and the agent asks whether you want a limited platform and scanner audit.

Will the skill tell me how many dollars I can save?

No. Its rules say to use cost magnitude phrases only and never customer-facing dollar savings figures. Recommendations are tied to observed metrics, verified files with line numbers and citations from a curated, version-aware docs list.

More in Skill directory

All guides

Keep your prompts and skills in one place, in your own GitHub. Install every skill with one command.

Get started

Kipt is an independent product, not affiliated with or endorsed by Anthropic, OpenAI, Cursor or GitHub. Claude, ChatGPT, Codex, Cursor and GitHub are trademarks of their owners.