googlarz/vibe-coder-kit/vibe-launch

vibe-launch

Overview

vibe-launch is the pre-production checklist for solo vibecoders. Run it before sharing your app with real users for the first time — or before any significant public moment: Product Hunt, a social post, sending a link to a client.

The problem it solves: vibecoders often "launch" by pasting a URL into a message before checking if the basics are safe. This skill makes you pause for 10 minutes so you don't spend the next 3 hours putting out fires.

Run it once. Then go ship.

Before you start

Ask the user:

"Before we run through the checklist — what's the live URL? And who are you about to share it with?"

This frames everything that follows. A link to a friend is different from a Product Hunt post.

The 6 Checks

Work through each check one at a time. Ask the user to confirm before moving to the next. This is a conversation, not a form — don't present all six checks at once.

Run these in order. Present each as a section. Wait for the user's input where needed. Do not skip any check.

Check 1 — Secrets safe?

This is the only hard blocker. A leaked API key or exposed database password can cost real money or compromise real users within minutes of going public.

Do all three:

1. Check for exposed secrets in the codebase. Run /vibe-check. If vibe-check isn't set up yet, I'll scan silently for common patterns — files containing sk-, AIza, AKIA, Bearer , database connection strings with passwords, or anything that looks like password=, secret=, api_key= with a real value next to it. You don't need to read code for this.

1. Confirm `.env` is in `.gitignore`. Run: cat .gitignore | grep .env or ask the user to check. If .env is not listed there, add it before doing anything else.

1. Confirm the git history is clean. Ask: "Has your .env file ever been committed to git by accident?" If yes, the key is already exposed — it must be rotated, not just deleted from the latest commit.

Result:

• ✅ Safe — no exposed secrets found, .env is gitignored, history is clean
• 🚨 Fix before launching — leaked secrets found; do not proceed until rotated and removed

If there's a 🚨, stop here. Help the user fix it. Do not continue until Check 1 passes.

Check 2 — Is it actually deployed?

A surprising number of "launches" share a URL that returns a 404, shows a Vercel build error, or still points to localhost.

Ask the user to verify in an incognito/private browser:

"Open this URL in a private browser window — not your regular browser where you're logged in. Does it load? What do you see?"

Incognito is important: your normal browser has cached versions and active sessions that can hide broken deployments.

Then ask:

• "Is this using your production environment variables — not your local .env?" If they're on Vercel, the env vars need to be set in the Vercel dashboard, not just in the local .env file. A common failure: the app deploys but can't connect to anything because the keys are missing.

• If the app uses a database: "Is it pointing at the production database, not a local or test one?"

Result:

• ✅ Live — URL confirmed working in private browser, production env confirmed
• ⚠️ Not confirmed — user couldn't verify, or env variables not checked

If there's a ⚠️, help them resolve it before continuing.

Check 3 — Does the core thing work?

The app might be "live" but broken in ways you haven't noticed because you've been testing the same account for weeks.

Ask the user:

"What's the one thing a new user comes to your app to do? Describe it in one sentence."

Then walk through it together, step by step, as a brand-new user. The user must do this test — you can't access the browser. Guide them:

1. Open the live URL in an incognito window (not logged in, fresh state)
2. If there's signup: try signing up with a fresh email right now
3. Complete the main action from start to finish
4. Check: any blank pages, error messages, or buttons that do nothing?

Also ask: "Open it on your phone. Does it look usable on a small screen?" Many vibecoders only test on desktop — a broken mobile layout on launch day is a common surprise.

Result:

• ✅ Core flow works — new user can sign up and complete the main action; mobile looks usable
• 🚨 Core flow broken — this is a launch blocker. Do not share the link until the main thing works. Help the user fix it, or if it's complex, suggest running /vibe-oops.
• ⚠️ Minor issues found — something non-critical is broken (a secondary page, an edge case). State what it is and ask: "Is this something users will hit on their first visit? If yes, fix it first. If no, you can launch and fix it after."

Check 4 — What happens if it breaks?

This check is awareness, not a blocker. Most solo vibecoders find out their app is down when a user messages them — or they don't find out at all.

Ask:

"If your app goes down at 2am, how will you know?"

Explain the options in plain English:

• Vercel has basic deployment status and some uptime visibility in the dashboard — not full alerting, but better than nothing
• Sentry (free tier) catches runtime errors and sends you an email when something breaks — for full setup steps, run /vibe-monitor
• UptimeRobot (free) pings your URL every 5 minutes and emails you if it goes down

If they have none of these, this isn't a blocker — but basic uptime monitoring takes 2 minutes to set up and is free. For full setup steps, run /vibe-monitor. For now, verify you have at least one uptime check configured — check your UptimeRobot (or equivalent) dashboard and confirm it's monitoring your production URL. If you need to set it up from scratch, run /vibe-monitor after launch.

Result:

• ✅ Has alerting — some form of error tracking or uptime monitoring is in place
• ⚠️ No alerting — not a hard blocker, but: "You'll find out about problems when users tell you, which means things can break silently. Worth setting up before you share widely."

Check 5 — Can you communicate with users?

If something breaks after launch, can you reach the people who signed up?

Ask:

"If you need to tell your users 'the app is down, we're fixing it' — can you do that? Do you have their emails?"

And:

"If you need to take the app down for an hour to fix something, is there any way to let people know?"

Common answers:

• Email collection at signup: ✅
• A Twitter/X or Instagram account for the project: ✅
• Nothing: ℹ️

This is not a blocker for a first launch. But it becomes important fast once real people are using the app.

Result:

• ✅ Can reach users — email list, social account, or some communication channel exists
• ℹ️ No way to reach users — not a blocker, but worth setting up: "If something breaks, your users will just leave quietly and you won't be able to explain what happened."

Check 6 — What's your rollback plan?

If the first deploy after launch breaks something, can you get back to a working state quickly?

Ask:

"Do you have a recent git commit you could go back to if things went wrong?"

And cover the three common scenarios:

1. Git checkpoint: I'll run git log --oneline -5 and show you the last few commits — we'll confirm a clean checkpoint exists. If not, I'll create one: git add -A && git commit -m "checkpoint before launch"

1. Database: If the app has a database, ask: "Do you have a backup?" For Supabase: Dashboard → Settings → Database → Backups. For Railway/Render: check the dashboard for backup options. For local Postgres: I'll run the backup command for you.

1. Vercel rollback: If they're on Vercel, they can roll back instantly — go to the Vercel dashboard, find the previous deployment, click "Redeploy." Takes 60 seconds. Make sure they know this exists.

Result:

• ✅ Can roll back — git checkpoint exists, Vercel rollback understood
• ⚠️ No rollback plan — help them create a git checkpoint before launching; it takes 30 seconds

Final Verdict

After all 6 checks: if any check is 🚨, don't launch. If all checks are ⚠️ or better, you can launch with awareness. If all checks are ✅, you're clear.

Once you've delivered the verdict, record that the checklist ran so the kit stops nudging about it: if a .vibe/ directory exists, run touch .vibe/.launched.

Edge cases

"I just want to share it with one friend to test" Still run Check 1 (secrets) and Check 2 (is it live). Skip or abbreviate 3–6. Even one person sharing a link can make a URL public.

"I've already launched, I'm doing a Product Hunt post" Run all 6 checks. A Product Hunt post can send thousands of people to your app in an hour. Check 3 (core flow) is especially important — test it fresh, not from memory.

"The app doesn't have a database / doesn't have users" Skip the database parts of Check 6 and the communication parts of Check 5. Note it explicitly so the user knows you're not forgetting it.

Verification checklist

• [ ] Check 1 passed: no secrets in code, .env is in .gitignore, git history clean
• [ ] Check 2 passed: deployment confirmed working on real device
• [ ] Check 3 passed: core user flow tested end-to-end
• [ ] Check 4 reviewed: either uptime monitor is set up, or ⚠️ noted in the final verdict
• [ ] Check 5 passed: contact/support method exists for users
• [ ] Check 6 passed: rollback path confirmed (know how to undo the deploy)
• [ ] Each check presented conversationally — user responded before moving to next