Goldie makes App Store screenshots from a coding agent, with 1,633 stars

Goldie is an MIT-licensed tool by Kacper Kapuściak, sponsored by Software Mansion, that generates App Store and Google Play screenshots and preview videos for mobile apps. It is at 1,633 stars over a 30-day window and integrates as a skill or plugin for Cursor, Codex and Claude Code, so a coding agent can produce store assets as part of a release.

What it does end to end

The pipeline starts on an iOS simulator or an Android emulator. Goldie replays a flow you defined, captures screenshots as it goes, wraps them in device bezels and backgrounds, and joins the clips into a preview video. Before it returns anything it checks the outputs against the store's upload rules, so a shot that is the wrong resolution, the wrong aspect ratio, or missing a required device type fails the run rather than the submission.

The README lists SwiftUI, UIKit, Jetpack Compose, Flutter, React Native and Kotlin Multiplatform as supported. That covers most mobile frameworks a small team is likely to use.

Why the agent integration matters

A coding agent that already works in the mobile codebase is well suited to run this. It can update the flow definition when a screen changes, re-run the capture, and open a pull request with the new assets, without a designer having to redo a set that the store would reject after a small change. That is what differs from other screenshot tools, which run outside the coding tools and rely on a person to trigger them.

Sponsorship by Software Mansion, a React Native consultancy that maintains a number of the popular RN libraries, is worth noting for one reason: it means a company whose main business is building mobile apps is behind the maintenance, rather than a solo author who may lose interest.

If store assets are a chore your team keeps missing on release week, Goldie moves the work to a place the agent already runs. The install is a plugin for whichever agent you use. The cost of getting it wrong is a set of screenshots you would have needed anyway, produced faster. The cost of getting it right is that a step which used to block submission no longer blocks it.

Two things to plan for. First, the flow definition is the part that goes out of date: a change to onboarding or a new subscription screen means the flow has to change too, so it belongs in the codebase and in code review. Second, the store compliance checks are Goldie's understanding of the current rules, not a substitute for the App Store Connect and Play Console rejection messages, so a first submission is still the fastest way to find out what the real requirements are.

Source

Primary source: kacperkapusciak/goldie on GitHub.