mrkeyoor.com_
Wed 16 Sept 06:42 UTC
Automationevaluationupdated 16 Sept 2026

goldie review

Goldie is a command-line tool and coding-agent skill that turns scripted mobile-app flows into framed App Store and Google Play screenshots. It also records preview clips, opens a browser studio for visual edits, and checks the exported media against store rules.

Verdict

Our Goldie install took 94 seconds and 492 MB, and its 97-second build passed, but there was no test target to run. It is worth trying when store media is recurring work and your team already maintains simulator flows. Skip it for a one-off screenshot deck or when nobody can own Xcode, Android emulators, and flow repairs.

We ran it

Lab card: what happened when we ran goldieScreenshot of goldie (github.com/kacperkapusciak/goldie)
Install✓ · 94s219 packages · 492 MB
Build✓ · 97s
Testsn/ano test script
Repo89 files~7,901 lines of source · 19.7 MB · 0 CI workflows

Answers from our run

Does goldie build from source?

Dependencies installed in 94 seconds (219 packages), and the build succeeded in 97 seconds. We cloned commit 788dd6e into a clean Debian container with 3 CPUs and no project-specific setup.

Does goldie have tests you can run?

Not through a standard command: the project exposes no test script or target that our harness could run.

Who should not use goldie?

iOS teams without a Mac and Xcode: App Store captures require macOS with iOS simulators.

What are the alternatives to goldie?

fastlane, App Store Screenshots, Koubou. Our Goldie install took 94 seconds and 492 MB, and its 97-second build passed, but there was no test target to run.

Setup3/5Build passed, but capture adds ffmpeg and a simulator or emulator
Docs4/5Clear commands, device rules, outputs, and platform caveats
Community3/52,076 stars and recent activity, but only a short public history
Maturity2/5No test target or CI workflows in the measured checkout

Who it’s for

Mobile teams that repeatedly produce store screenshots across devices or locales.
Developers willing to describe app journeys as Argent flows and repair them when the UI changes.
Coding-agent users who want the agent to create the flows and configuration, then leave editable files behind.
Designers who want a browser studio backed by the same design file the CLI renders.

Who it’s NOT for

iOS teams without a Mac and Xcode: App Store captures require macOS with iOS simulators.
Developers looking for a browser-only screenshot decorator: Goldie drives a built app through simulators or emulators and depends on Argent flows.
Apps whose interface changes often without anyone owning automation maintenance: the README says flows fail when the app changes.
Teams requiring an established test and CI record before adoption: our checkout had no test target, no tests directory, and 0 CI workflow files.
Anyone unwilling to install the platform toolchain: Node 20 or newer and ffmpeg are baseline requirements, with Xcode or an Android emulator added for capture.

Setup reality

Our sandbox installed Goldie in 94 seconds, adding 219 packages and using 492 MB. The build succeeded in 97 seconds. There was no test script or target, so the lab skipped tests rather than reporting a pass.

Node 20 or newer and ffmpeg must be on the host. App Store work requires macOS and Xcode simulators. Google Play capture needs an Android emulator; Linux also needs Argent's extra emulator packages. Your config points scenes at Argent flows and names the app build.

Use a release build or debug banners can appear in the captures. Android needs a Pixel 10 Pro or Pixel 9 Pro hardware profile. Goldie can boot a matching virtual device, but flows still need selectors that work on each platform and must be repaired when the app changes.

Goldie automates the capture as well as the frame

Goldie uses Argent to replay flows inside an iOS simulator or Android emulator. It captures the app, places each image in a device bezel, adds a background and headline, and can join recorded clips into a store preview. That chain is the reason to consider it. Plenty of tools decorate images you already made. Goldie tries to own the repeatable route from app state to upload-ready media.

The output is specific. Its iPhone preset makes 1320 x 2868 screenshots and an 886 x 1920 H.264 preview, while the Google Play phone preset produces 1080 x 1920 images. Apple's preview must last 15 to 30 seconds. Android preview output is intended for YouTube, so Apple's duration rule does not apply there. Those baked-in dimensions remove a tedious source of rejected uploads.

What happened when we ran it

Our sandbox install finished in 94 seconds. Bun added 219 packages and the checkout occupied 492 MB after installation. The build then passed in 97 seconds. The repository itself was much smaller, with 89 files, about 7,901 lines of source, and a 19.7 MB checkout at commit 788dd6e.

There was no test script or target for the lab to call, so tests were skipped. That is different from a passing suite. The scan also found 0 CI workflow files, no Dockerfile, and no tests directory. None of those absences proves the generated screenshots are wrong. They do mean our successful compile is the strongest automated result from this run, not evidence that simulator replay or store verification worked end to end.

App Store capture still requires a Mac with Xcode

Node 20 or newer and ffmpeg are the baseline. For iPhone screenshots, Goldie requires macOS, Xcode, and an iOS simulator. Android capture works on macOS, Linux, or Windows through an emulator. Linux hosts need the extra packages named in Argent's emulator prerequisites, and a non-Mac config should keep only the Android device preset.

The Android path has its own exact requirement. Goldie expects an AVD based on the Pixel 10 Pro or Pixel 9 Pro hardware profile, both with a 1280 x 2856 screen. The configuration also needs the APK path and application ID. A running matching emulator can be reused, or Goldie can boot it. Selectors must still find the right controls on both platforms if you want one flow to serve each store.

The agent skill leaves maintainable files behind

The agent workflow begins inside the app repository. Goldie asks which stores you want, explores the app, writes Argent flows and configuration, then opens its studio. Later prompts can change the same files rather than create another disposable image. Installation instructions cover the shared skills format plus a Claude Code marketplace plugin, which makes claude-code a relevant tag rather than a marketing guess.

Automation does not make the flow permanent. The README says flows fail when the app changes and suggests repairing or rerecording them with Argent. Use release builds because debug builds can paint LogBox banners into the capture. If screenshots are produced every release or in several locales, maintaining those flows can pay back. For six images made once, the setup may take longer than manual capture.

One design file drives the studio and final render

The browser studio writes its choices to goldie.design.json, and the CLI renders from that same state. You can change devices, backgrounds, layout templates, bezels, fonts, and copy per tile. The repository includes six font families and several iPhone frame colors, while Android can use the bundled Pixel bezel or custom frame art. Decorations sit behind the device rather than forcing edits into the raw capture.

The command split is sensible: doctor checks tools and flows, all captures and renders, and studio handles visual adjustment. Finished assets land in separate screenshot and preview folders by device and locale. Verification against store rules is useful, though our 97-second build did not exercise a simulator, emulator, or upload. Treat doctor and one complete device run as the acceptance check on your own host.

September activity comes with a short track record

GitHub says the repository was created on August 24, 2026, and last pushed on September 2. It had 2,076 stars and 4 combined issues and pull requests when fetched. The package manifest reported version 0.3.1 and an MIT license, while GitHub's latest-release endpoint returned no published release. That describes a young project with strong early attention, not a settled release history.

Two open reports document Linux and Windows installation trouble in version 0.3.0. The current package manifest now lists macOS, Linux, and Windows, and the README explains what each host can capture. Those facts prevent us from presenting the older reports as confirmed current breakage. They still show why platform checks belong in your trial, especially outside the Mac and iOS path.

Goldie pays off when you will rerun the workflow

A 492 MB install plus simulator tooling is hard to justify for one launch graphic. The equation changes when each release needs fresh captures, several device sizes, or multiple locales. Goldie records the app journey, design choices, and output rules as files you can rerun. Its missing test target keeps it in trial territory, so make one real release build traverse every intended flow before letting the generated deck replace a manual review.

Alternatives

ProjectWhat it isPick it when
fastlane gh↗A broad mobile release automation suite that includes screenshot generation and store delivery.pick this instead when screenshots belong inside a larger signing, metadata, and release pipeline.
App Store ScreenshotsAn agent skill that creates a connected-canvas editor from supplied app captures.pick this instead when you already have source images and want a visual editor without simulator replay.
KoubouA YAML-driven screenshot generator with HTML templates, localization, and live previews.pick this instead when deterministic template rendering matters more than automated app navigation or preview video.

What people are saying

  1. [velocity-scout] kacperkapusciak/goldie

Sources

  1. Goldie README
  2. Goldie package manifest
  3. Goldie repository
  4. Goldie Windows studio issue
  5. Goldie Linux installation issue
  6. Goldie commit tested by MrKeyoor

More automation reviews

TikTokDownloader · goreleaser · obscura · ios_rule_script · background-agents · firstmate · the whole board →