Everyone agrees an app should be production-ready before it launches. Ask what that means and you get a shrug, or a list of two hundred items nobody reads.
Ours has twelve. They fall into four groups, and each one is a question with a yes or no answer. If an app passes all twelve we are happy to launch it. If it fails one, we know exactly what to do next.
The four groups
| Group | The question behind it |
|---|---|
| Safe | Could this hurt the people who use it? |
| Sound | Will it stay up, and come back if it falls over? |
| Tested | Do we know it works, without clicking through it by hand? |
| Maintainable | Can someone else change it next year? |
Safe
- 01No secrets in the codeKeys and passwords live in a vault or in the host's settings, never in the repository or the browser.
- 02Every route checks who is askingA signed-in user can only reach their own data, and the server decides that, not the screen.
- 03Input is never trustedAnything typed, uploaded or sent is checked and limited before it is used or stored.
These come first because they are the ones that harm other people. A slow app is embarrassing. A leaked customer list is a legal matter.
Sound
- 01Backups that have been restoredA backup counts once someone has restored from it. Until then it is a hope.
- 02Errors reach a personWhen something fails, an alert goes to someone who can act, before a customer writes in.
- 03It copes with a bad dayA payment provider or an AI model being down gives a clear message and a retry, not a blank page.
Tested
- 01The money paths are coveredSign-up, sign-in, checkout and anything that writes data have automated tests.
- 02Tests run on every changeNothing reaches users unless the tests have passed. No exceptions for small fixes.
- 03It is tested in a real browserOn a phone, a tablet and a desktop, the way a visitor would use it.
This is the group AI-built apps most often skip entirely, and it is the one that pays back fastest. Tests are what let you change the app next month without fear.
Maintainable
- 01One way of doing each thingThe same job is not done three different ways in three different files.
- 02Releases are one repeatable stepAnyone on the team can ship, and anyone can roll back.
- 03It is written downHow to run it, how to release it and where things live, in the repository, kept current.
How to use the list
Go through it honestly and mark each one yes or no. "Mostly" is a no.
- 01Score itTwelve questions, yes or no. It takes an afternoon.
- 02Fix Safe firstAnything in the first group is fixed before a single new feature.
- 03Then TestedTests go in next, so every later fix can be proved.
- 04Then the restSound and Maintainable follow, in whatever order hurts most.
A no is not a failure. It is the next piece of work, already described.
You do not need all twelve to show the app to five friendly customers. You do need all of Safe. For a public launch, with payments or personal data, we would want all twelve.
Where this comes from
It is the same standard we hold our own software to. The platform our shop runs on will not release unless its tests pass in a real browser on desktop, iPad and iPhone, and you can read how that works in the Eskdale Platform case study.
If you would like your app scored against the list by someone who did not build it, that is the first step of our vibe-coded app service.
