MaarkssDN/Birthub-360 / .agents/skills/appium-testing
appium-testing
VerifiedUse whenever one mobile UI test suite must run on iOS and Android (native or React Native apps), or an Appium suite fails, flakes or breaks after an upgrade: XCUITest, UiAutomator2 and Espresso drivers, capabilities, clients in JS, Python, Java, Ruby or .NET, simulators, emulators and device farms.
evals/shared-suite-only-runs-on-android/criteria.json
View on GitHub{
"context": "Two skill-specific facts decide this task. First, feature coverage is per automation backend, not per framework: `startActivity`, `pressKeyCode`, `toggleWiFi`, `openNotifications` and `mobile: shell` are UiAutomator2-side commands with no XCUITest counterpart, so a shared suite either uses a cross-platform substitute (a deep-link launch through the URL-handling capability, an in-app back control, an in-app assertion instead of a shell read) or declares the test platform-limited. Second, this framework is an out-of-process HTTP server: every element lookup and every getText is a round trip costing on the order of a hundred to a few hundred milliseconds, which is why a 200-iteration tight loop is unnoticeable against a local emulator and costs minutes against a farm on the other end of a network. Predicted baseline failure: the agent wraps the failing calls in try/catch or an `if (isAndroid)` guard that returns early - leaving six tests green and empty, which is what the team already has - and never explains which backend supports what; and it leaves the sync helper alone, or 'optimises' it by adding a fixed pause inside the same unbounded loop, because a loop with no sleep in it reads as fast. Both failures are invisible on a dashboard that still says 22/22.",
"type": "weighted_checklist",
"checklist": [
{
"name": "Failing calls identified as backend-specific with a per-platform verdict",
"max_score": 28,
"description": "`docs/platform-support.md` goes call by call - activity launch, keycode press, wifi toggle, notification shade, device shell - and states for each whether the Android backend and the iOS backend support it, with the substitute where one exists. A document that lists the tests without naming the backends, or that asserts blanket cross-platform support, scores zero. Covering only some of the five calls scores at most half."
},
{
"name": "Shared paths use commands both backends implement",
"max_score": 22,
"description": "Where a cross-platform route exists it is taken - launching the deep link through the session's URL handling rather than an activity name, using the app's own back control rather than a platform keycode, asserting the breadcrumb through the app UI or an app-exposed hook rather than a device shell read. Wrapping the original call in try/catch and continuing scores zero. Replacing an Android-only call with a different Android-only call scores zero."
},
{
"name": "Polling helper reduced to a bounded wait",
"max_score": 22,
"description": "`waitForSync` becomes a wait with an explicit timeout and an explicit polling interval (the client's own wait-until primitive, or an equivalent bounded loop with a deadline and a delay), so the number of round trips is a function of elapsed time rather than a fixed two hundred. Leaving the tight 200-iteration loop scores zero. Keeping the 200-iteration bound and adding a fixed pause inside it scores at most half - it costs less but still has no deadline and still hard-codes the iteration count."
},
{
"name": "Slowness explained as protocol round-trip cost",
"max_score": 14,
"description": "The answer states that each element lookup and text read is an HTTP command to an out-of-process server, costing on the order of 100-300ms and considerably more against a remote farm, so a fixed 200-iteration loop is hundreds of round trips. Explaining the difference as 'the farm devices are slower' or 'the network is slow' without naming the per-command round trip scores zero; no explanation scores zero."
},
{
"name": "Platform-limited tests are reported as not run",
"max_score": 14,
"description": "Any test that genuinely cannot run on iOS is excluded through the runner's skip mechanism with a stated reason, so the report shows it as skipped. An early `return` inside the test body, a guard that leaves the test green with no assertions, or leaving the external skip list as the mechanism scores zero - the task states that a silently green test is worse than a red one."
},
{
"name": "MUST NOT split the suite per platform",
"max_score": 12,
"description": "MUST NOT. Anti-pattern row 'Per-platform separate test code'. Moving the six tests into an android-only spec directory, or creating an iOS twin of any spec, scores zero even though it would make the run green - the task states the merged suite is being kept."
},
{
"name": "No test deleted and assertions preserved",
"max_score": 8,
"description": "All six tests still exist and still assert what they asserted for the platform that supports them. Deleting a test, or reducing it to a launch with no assertion, scores zero."
}
]
}
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |