Launch Readiness Checklist for Vibe-Coded Apps

Work through 35 pre-launch checks an AI builder skips: row-level security, exposed API keys, backups that restore, failure cases, alerting and getting your code out.

The interactive tool loads here when JavaScript runs.

About this tool

What is the Launch Readiness Checklist?

An app built with Lovable, Bolt, v0, Replit Agent or Cursor usually works. The question is whether it is safe to put in front of people, and that is a different question. This checklist covers the 35 things that are not features: who can read the data, where the API keys ended up, whether a backup actually restores, what the app does when the network drops mid-submit, and who finds out when it breaks at two in the morning.

Pick your stack and how deep you want to go, then work down the list. Each item says what goes wrong in plain terms and what to do about it, concretely enough to act on without a tutorial. Nothing is sent anywhere: the list runs in your browser, your ticks stay on your device, and you can copy or download the result as Markdown to paste into an issue tracker.

  • 35 checks across eight themes, from data access to legal disclosure
  • Filter by stack: Supabase, Firebase, another hosted database or serverless functions
  • Filter by risk, so critical items can be worked first
  • A counter that tracks critical items separately from the total
  • Markdown export with unchecked items first and their "what to do" line attached
  • Runs entirely in your browser; ticks are never uploaded
Read the guide: What to Check Before Launching an App You Vibe-Coded

Updated

How to use the Launch Readiness Checklist

  1. Choose your stack. Supabase, Firebase, another hosted database or serverless functions each bring a few items that do not apply to the others; choose Everything to see the full list.
  2. Choose how much to show. Critical only is the short answer to "can I launch this week". Everything, including medium is the thorough pass.
  3. Work down the themes in order. Data access and secrets come first because they are the two that actually hurt people.
  4. Tick items as you confirm them, not as you plan them. The counter tracks critical items separately, so you can see at a glance whether anything launch-blocking is left.
  5. Copy or download the remaining work as Markdown. Unchecked items come first and carry their "what to do" line, so the export is a worklist rather than a record of what you already did.

Worked examples

The week before a launch

Input

Stack: Supabase. Depth: Critical only

Output

9 critical items across data access, secrets and backups

The shortest list that still answers whether launching is safe.

Handed someone else's app

Input

Stack: Everything. Depth: Everything, including medium

Output

All 35 items, grouped into eight themes

The full first pass when you did not build it and need to find out what you have inherited.

After a scare

Input

Tick everything confirmed, then download

Output

launch-readiness.md, unchecked items first

Paste straight into an issue tracker so the remaining work is visible to the whole team.

Background and insight

AI builders are excellent at the happy path and quiet about everything else. They will produce a working signup, a working dashboard and a working checkout, because those are the things a prompt describes. Nobody prompts for "and make sure a second user cannot read the first user's rows", so that part is left at whatever the default was.

The default is the problem. An app generated this way usually talks to its database straight from the browser using a key that ships inside the bundle, which is safe only because row-level security is supposed to stop that key from reading anything it should not. When row-level security is off, or on but missing a policy for one of the four operations, the interface looks completely normal and the data is readable by anyone who opens the network tab.

The second pattern is secrets. A payment or model-provider key pasted into a component ends up in the JavaScript every visitor downloads. Removing the line later does not help, because the key is already in git history and in the bundles people have. It has to be rotated at the provider and moved behind a server function.

The rest are the ordinary operational things that any application needs and that no prompt produces: a backup you have restored at least once, error cases that fail visibly rather than hanging, a way to learn about a failure before a customer tells you, and a copy of the code somewhere the platform does not control.

Common use cases

  • Before a first launch

    Find the data-access and secret problems while the user count is still zero.

  • Taking over an app

    A repeatable first pass when a client or a founder hands you something built in a builder.

  • After the app misbehaves

    Work out which operational gap produced the incident, and close the rest before they do the same.

  • Deciding whether to depend on a platform

    The portability items turn vendor lock-in from a feeling into a short list of facts.

Tips and limitations

  • Test access with a second account, not your own. When you are the only user every row belongs to you, so a cross-account leak cannot appear.
  • Row-level security being enabled and a policy existing are two different settings. A table with policies but RLS switched off is still open.
  • A key that was ever in a frontend bundle or a public repository has to be rotated, not deleted. The copies people already downloaded do not change.
  • A backup you have never restored is a file you hope is a backup. Restore it once into a scratch project.
  • Take an export before letting an agent change the schema. A prompt that drops a column looks exactly like a prompt that adds one.
  • Set a spend cap on anything billed by usage. A loop or a scraper turns a quiet week into a large invoice.

Sources and further reading

  1. Row Level Security, Supabase. Enabling RLS and writing a policy per operation
  2. Using Row Level Security, PostgreSQL. The database behaviour the hosted guides build on
  3. Secrets Management Cheat Sheet, OWASP. Why a rotated key is the only fix for an exposed one
  4. OWASP Top Ten, OWASP. Broken access control is the most common serious web weakness
  5. Web Content Accessibility Guidelines (WCAG) 2.2, W3C. The keyboard and contrast items

Frequently asked questions

Is an app built with an AI builder safe to launch?

It can be, but not by default. The generated features usually work; what is missing is the configuration around them, chiefly row-level security on every table and keeping secret keys out of the browser bundle. Those two account for most of the real harm, and both can be verified in an afternoon.

What is row-level security and why does it matter so much here?

It is the database rule that decides which rows a given user may read or change. Apps built this way query the database directly from the browser, so the database itself is the only thing standing between one user and everyone else's records. With it off, the key in your bundle can read the whole table.

I pasted an API key into a component. Is deleting it enough?

No. The key is already in your git history and inside every bundle that has been downloaded since. Rotate it at the provider so the old value stops working, then call that service from a server function instead, with the new key stored as a secret there.

My platform says it backs up automatically. Do I still need my own export?

Take one anyway, and restore it once. Automatic backups have a retention window and live in the same account as the thing they protect. A restore you have actually performed is the only way to learn that the export is complete and usable.

Does this checklist send my answers anywhere?

No. The list, your ticks and the Markdown export are all produced in your browser, and nothing is transmitted or stored on a server. Closing the tab clears the ticks, which is why the export exists.

Is there an in-depth guide to the Launch Readiness Checklist?

Yes. “What to Check Before Launching an App You Vibe-Coded” covers the background, how it works, common pitfalls and worked examples. Read the full guide.