Audit your own app's screens for drift

Reverse every screen you ship, read them side by side, and find the places your interface stopped agreeing with itself.

Inconsistency arrives one merged pull request at a time and is invisible to the person who added it. Reading forty screens at once is the only view in which a second button style is obvious.

Why this will not look generic: The findings come from counting your own shipped screens, so the inconsistencies are ones your users are actually looking at rather than ones a linter can see.

What you walk away with

• A prompt per screen, in one folder
• The patterns your interface genuinely follows
• The screens that break them, named
• A short list of what to reconcile first

Typical run: 20 to 80 screens. Capture at one viewport width. Mixing mobile and desktop shots makes every difference look like drift.

This one happens in one conversation

You bring: Screenshots of every screen you ship, captured at one consistent width.

Open with: These are all screens from one app. What visual rules does it actually follow, and which screens break them?

What the thread holds on to:

• Every screen at once, which is the only way a second button style becomes visible
• The rules it inferred, which you can test by asking it to predict an unseen screen
• Which breaks you have already said are intentional

Then you can just ask:

• Which of these would a user actually notice?
• Is the outlier screen wrong, or is it the newest and the rest are stale?
• Give me the three fixes with the best ratio of consistency gained to work

What the reverse actually gives you

Point it at: Forty screens of your own product, captured at one width.

What comes back (representative example): The system is mostly consistent and breaks in two places. Thirty-six of forty screens use the same card treatment: one-pixel border, no shadow, generous internal padding. Four use a shadowed card with no border, and all four are in the settings area. Primary buttons are a solid fill everywhere except the onboarding flow, which uses an outline style that appears nowhere else in the product. Spacing, type and palette are consistent throughout. The two breaks correlate with areas rather than with dates, which points at them having been built by someone working from a different reference.

Things you would never have thought to type:

• shadowed cards in exactly four screens, all in settings
• an outline button style that exists only in onboarding
• the breaks correlating with area, not with age
• spacing and type consistent, which narrows it to components

What you change: Two component usages, not forty screens. The value is that the audit tells you it is a components problem rather than a general tidiness problem.

How it works

1. Reverse every screen (In your code) - All of them, not the ones you are proud of. The drift lives in the screens nobody revisits.
2. File them as one set - One folder. Tag by area if you want the answer to tell you where the drift clusters.
3. Ask what the rules are, and who breaks them (In Claude) - Test the inferred rules by asking it to predict a screen it has not seen.

Cost and time

This workflow runs at 2 reverses + about 1 minute. Your result carries automatically from each step into the next, so there is no re-uploading between tools. Reverses are billed from your reverse meter and generation steps from your AI Credits, both bundled in every pay-as-you-go credit pack.

Browse all guided workflows or all individual tools. Pay as you go, no subscription, AI Credits never expire.