Maestro Apple App Testing

Maestro Apple app testing

Use this as the main workflow for iOS app testing with Maestro. Combine it with the project's XcodeBuildMCP and Apple-development skills; this skill owns the end-to-end testing loop, not product implementation.

Inputs and output

Inputs: the active Apple app workspace, its Xcode project or package, available simulators/devices, existing .maestro/ flows, and the requested behavior or bug.

Output: repeatable Maestro flows or focused edits, executed test results, screenshots and debug artifacts when useful, and a concise report of coverage, failures, and remaining manual checks. Never claim a test passed without running it.

Scope Maestro to changed journeys

Do not run Maestro merely because code changed. Use it when a task edits a Maestro flow or changes user-visible UI, interaction, navigation, or journey behavior. For non-UI code, run Maestro only when the changed logic directly affects a known user journey; use focused unit, integration, or build verification for documentation, tooling, refactors, and logic outside a user journey.

Map the change to its existing flow(s) before testing. Run only the one edited or directly affected flow during normal work; if a shared change genuinely affects multiple journeys, run only those directly affected flows and state why. Do not run tagged suites or every flow as a routine final check. Run a complete suite only when the user explicitly requests it or an explicit release/CI policy requires it.

Workflow

1. Inspect the active workspace, existing .maestro/ flows, app identifiers, accessibility identifiers, and available test fixtures. Reuse existing conventions.
2. Check the toolchain with command -v maestro, maestro --version, and maestro list-devices. If Maestro or its configured MCP server is unavailable, report the exact blocker before writing speculative flows.
3. For Swift/Xcode builds, runs, simulator management, and UI inspection, use the available XcodeBuildMCP tools. Before the first build, run, or test action, call session_show_defaults and report the active project, scheme, configuration, and simulator. Resolve and reuse one exact simulator ID; do not use OS=latest.
4. Build and launch the app with the project's established XcodeBuildMCP workflow. Use Maestro MCP for live hierarchy inspection, screenshots, interaction, and debugging; use maestro test for repeatable YAML execution and CI-like coverage.
5. Author flows with stable testID/accessibility selectors, explicit assertions, automatic waiting, conditional handling for permissions/onboarding, and reusable subflows. Inspect system-generated controls such as SwiftUI tab buttons before selecting them: an identifier inside .tabItem may not propagate to the generated button, so use its confirmed accessibility label when no identifier is exposed. Do not use arbitrary sleeps when a state assertion or wait can express the condition.
6. When authoring or repairing a flow, cover its relevant happy path, validation and error states, loading/empty states, relaunch and persistence behavior, navigation/back behavior, permissions, keyboard/input, and accessibility-visible labels. Keep that coverage in the focused flow; do not expand to unrelated flows.
7. On failure, preserve or request debug output, screenshots, hierarchy, and video when useful; classify the failure as app behavior, test selector/flow, fixture, simulator/device, build, or environment. Reproduce once before changing a flow.
8. Finish by reporting exact commands, simulator/device ID, flows and tags run, pass/fail counts, artifact locations, and untested or manually verified areas.

Maestro MCP rules

The official Maestro CLI includes the MCP server. Prefer the configured local server with maestro mcp for interactive debugging. Use MCP to inspect the current screen and confirm selectors before editing a flow; use CLI flows for deterministic regression coverage. Keep MCP and CLI pointed at the same simulator/device and app state.

If a task requires Android or web coverage, adapt the workflow to the active platform and report platform-specific gaps rather than assuming iOS selectors or behavior.

Verification contract

Run the single smallest relevant flow after each flow or journey change. Do not add a tagged or complete-suite run unless the user explicitly requests it or a documented release/CI gate requires it. Use --debug-output or equivalent artifacts for failures. Do not modify production data or use real credentials; use fixtures, test accounts, and environment variables supplied by the workspace.