More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Vercel Deployment: Production Best Practices

    1

    Deploy Next.js on Vercel and run it like production: preview-first promotion, per-environment variables without leaked secrets, right runtimes, and a tested rollback.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Vercel Deployment: Production Best Practices

    Vercel Deployment: Production Best Practices

    Example session with this skill installed

    We are launching our Next.js marketing site and app on Vercel next week. Give us the go-live plan.

    • Read your context and instructions
    • Compiled the vercel deployment

    GO-LIVE PLAN — produced by applying the skill.

    Setup (days 1–2): import the repo in the dashboard and verify the framework preset reads Next.js before the first deploy; pin the Node version to your local one. First deploys are previews only — production is not connected until the preview loop has run clean.

    Env matrix (day 2, verified line by line before launch): list every variable the app reads, mark each public-safe or secret. Anything the browser needs gets a public identifier only — the payments publishable key may carry the NEXT_PUBLIC_ prefix; the secret key, database URL, and session secret never do, and if any code path seems to need a prefixed secret, that is a code bug to fix, not a config choice. Preview values point at staging services; production values are entered separately, never copied across. Remember the budget: 64KB total per deployment, 5KB per variable at the edge.

    Promotion flow (days 2–5): PRs get preview URLs for QA and sign-off, protected if the app is not public-ready; required checks (build, typecheck, tests) gate merges; production deploys come from merges only — no direct pushes during launch week, especially not "just this once."

    Launch day: lower DNS TTLs the day before; configure apex and www, confirm HTTPS; verify production env against the matrix; enable analytics, speed insights, and failed-deploy plus error-rate alerts to a human. Then do one rollback and re-promote as a drill while stakes are zero, so the first real rollback is not also the first attempt.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Vercel makes the first deploy take five minutes — and makes it just as easy to ship a production site with a secret exposed to the browser, staging values in production, and a rollback path nobody has ever pressed. This skill covers the deploy and the operating discipline after it. You get the two first-deploy paths (dashboard import and CLI) with the framework-preset and Node-version checks that prevent the most confusing build failures; the three-environment model (preview as disposable QA, production from merges only, custom targets for staging) treated as three different promises; and the env-var rules that bite everyone once — the 64KB total budget, the 5KB edge limit, redeploys being required for variable changes, and the NEXT_PUBLIC_ prefix that publishes whatever it touches to the browser, with a decision rule for splitting public identifiers from real secrets in code instead of exposing them. Runtime choice is made per function with a decision table (edge for middleware and lean endpoints, Node for ecosystem work, and neither for long-running jobs that belong on a queue), and the CI/CD loop adds required checks so a green preview URL is never mistaken for a green build. Production hygiene covers analytics and speed insights from day one, log retention limits, error alerting, and DNS cutover timing. A worked launch example, a troubleshooting table for the failures you will actually hit, an env matrix template, a pre-launch checklist, and a readiness rubric with a hard floor on secret hygiene and recovery round it out.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for DevOps & Cloud.

    • It is Vercel-specific by design: the env, runtime, and promotion mechanics do not transfer literally to other hosts, though the discipline does.
    • Plan limits (timeouts, memory, log retention) change over time and by tier — the skill tells you which numbers to check for your plan rather than guaranteeing figures.
    • It does not cover workloads that should not be on serverless functions at all beyond identifying them and pointing them at a queue or worker platform.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 24 days ago

    Permissions

    No special permissions declared or detected

    Listed24 days ago
    Updated21 days ago

    What's inside

    Frequently Asked Questions

    Environment variables, until proven otherwise. Preview and production have separate variable sets, and a variable added or edited after the last production build does not exist in the running deployment. Compare the two environments line by line against an env matrix, then redeploy — the skill's troubleshooting table orders the rest of the usual suspects after that.

    No — the prefix publishes the value to every visitor; that is the most common self-inflicted secret leak on the platform. The skill's rule: split the credential. Browsers get a public identifier (a publishable key designed to be seen); the real secret stays server-side, and code that seems to need the secret in the browser gets fixed instead.

    It is one click you have never pressed, during an incident, under time pressure, possibly by whichever teammate is on call. The drill costs five quiet minutes at launch and converts an unknown into a known — the worked example in the skill has the week-three bad-config rollback taking minutes precisely because the path was rehearsed in week one.

    New in DevOps & Cloud