vocello-ui-autonomy
Skill388 starsChanged 21 days agoMITView on GitHub
Run the default autonomous route for device and UI checks XCUITest does not cover — iPhone through iPhone Mirroring with computer use, the Mac directly, and Claude's review of lane screenshots — inside a maintainer-announced window.
In this repository8 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- website/CLAUDE.mdCLAUDE.md
- vocello-benchmarkSkill
- vocello-debugSkill
- vocello-device-diagnosticsSkill
- vocello-ios-validationSkill
- vocello-macos-uiSkill
- vocello-release-readinessSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 11
Read docs/reference/development-workflow.md (Computer use) and the routing table in docs/reference/ios-device-testing.md.
- Distinctive: line 44
Never drive the phone or the Mac while a native, device or measured lane runs.
- Very specific: line 64
After every XCUITest lane, read <run>/screenshots/index.json and open each failure frame and each named state the row asserts: look for …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Autonomous UI checks beyond XCUITest
Goal: $ARGUMENTS
Read docs/reference/development-workflow.md (Computer use) and the routing table in docs/reference/ios-device-testing.md. Maintainer decisions of 2026-10-25: XCUITest stays the only automated evidence; computer use is the default route for everything it cannot cover; a computer-use record closes an attended row only when the maintainer confirms it; phone work happens only in a window the maintainer announced; Claude never changes a system or security setting, grants a permission, enters a passcode or credential, deletes the app or confirms a purchase.
Route each check
- XCUITest can drive it (app UI with stable identifiers): use
scripts/ui_test.sh(skillvocello-ios-validationorvocello-macos-ui), then review the run's screenshots (below). - XCUITest cannot, and the check survives Mirroring's conditions (UI flows, in-app settings, model management, TestFlight installs, reproductions, visual inspection): computer use.
- Maintainer by hand: anything needing an iPhone system setting (Wi-Fi off, airplane mode, Auto-Lock, Low Data Mode, Developer Mode, UI Automation), a permission prompt, the passcode or Face ID, the microphone or camera, audio routing or interruptions (Mirroring sends audio to the Mac), VoiceOver, app deletion, a purchase confirmation. Pause and notify (below).
Window protocol (iPhone)
- Setup card, sent once at the window's start: Wi-Fi and Bluetooth on (both devices); Auto-Lock as the rows need; Developer Mode and UI Automation on; a Focus on so notifications stay out of screenshots; the build to test installed; Screen Recording allowed once for the capturing app; the phone unlocked.
- Block A, lanes:
scripts/ios_device.sh mirroring stop, then every XCUITest lane in scope. The lanes'ui-preflightrefuses while iPhone Mirroring runs, because XCUITest breaks under it. - Handoff: ask the maintainer to lock the phone (Mirroring needs it locked; unlocking ends Mirroring).
- Block B, Mirroring:
scripts/ios_device.sh mirroring status, thenscripts/ios_device.sh mirroring session <label>;request_accessfor "iPhone Mirroring";open_application"iPhone Mirroring"; screenshot before every action; act; after each meaningful stepscripts/ios_device.sh mirroring capture <label> --step "<action>" --observed "<what shows>"; verify before moving on. - Close:
scripts/ios_device.sh mirroring stop; report each record (folder under build/artifacts/ios/mirroring) for the maintainer's confirmation; never claim a row closed before it.
Never drive the phone or the Mac while a native, device or measured lane runs.
Input through iPhone Mirroring
- Tap: a click on the element. Scroll: scroll events over the Mirroring window, never a drag (a drag registers as a long press and opens Home Screen edit mode).
- Cmd-1 Home Screen, Cmd-2 App Switcher, Cmd-3 Spotlight, Cmd-4 Control Center (macOS 27).
- Typing has not reached the phone so far; until the first-window probe settles a method, avoid steps that need typed text or use the app's own paste from the Mac clipboard.
- A Mac authentication prompt for Mirroring, or any phone system sheet, is the maintainer's.
Mac
Drive Vocello and its menus, Finder reveal and drag-and-drop of reference files directly, between lanes only. No TCC grants, sign-ins or Sparkle installs. Report steps with screenshots; the attended walk in docs/reference/macos-release-qa.md is a typical session.
Pause and notify
At any step that is the maintainer's, stop, send a PushNotification naming the exact step, and resume only on their word. Keep the session and its record open meanwhile.
Review lane screenshots
After every XCUITest lane, read <run>/screenshots/index.json and open each failure frame and each
named state the row asserts: look for clipping, overlap, wrong state, untranslated text or an
unexpected dialog. Report-only; classify each finding as a release blocker or a backlog line.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

