Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Configure Flutter for desktop platforms with platform channels and native integrations
.claude/skills/a5c-ai-flutter-desktop-config/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -69% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 6% | 0% |
Configure Flutter for desktop platforms (Windows, macOS, Linux) with platform channels, native integrations, and platform-specific configurations.
json{ "type": "object", "properties": { "projectPath": { "type": "string" }, "platforms": { "type": "array", "items": { "enum": ["windows", "macos", "linux"] } }, "windowConfig": { "type": "object" }, "nativeChannels": { "type": "array" } }, "required": ["projectPath"] }
dart// Dart side class NativeService { static const platform = MethodChannel('com.example/native'); Future<String> getSystemInfo() async { return await platform.invokeMethod('getSystemInfo'); } }
swift// macOS (Swift) let controller = FlutterViewController() let channel = FlutterMethodChannel(name: "com.example/native", binaryMessenger: controller.engine.binaryMessenger) channel.setMethodCallHandler { call, result in if call.method == "getSystemInfo" { result(ProcessInfo.processInfo.operatingSystemVersionString) } }
cross-platform-test-matrixdesktop-build-pipeline process| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 10,781 | 9,522 | -12% | 1 | 1 | 0% | 2,304 | 2,188 | -5% | 0 | 0 | — |
case-02 | fail→pass | 14,628 | 12,420 | -15% | 1 | 1 | 0% | 3,253 | 2,668 | -18% | 0 | 0 | — |
case-03 | fail→pass | 15,811 | 3,334 | -79% | 1 | 1 | 0% | 3,517 | 1,096 | -69% | 0 | 0 | — |
case-04 | pass→pass | 13,146 | 10,727 | -18% | 1 | 1 | 0% | 2,821 | 2,643 | -6% | 0 | 0 | — |
case-05 | pass→pass | 22,378 | 17,752 | -21% | 1 | 1 | 0% | 3,744 | 3,906 | +4% | 0 | 0 | — |
case-06 | pass→pass | 14,401 | 12,012 | -17% | 1 | 1 | 0% | 3,150 | 3,024 | -4% | 0 | 0 | — |
case-07 | fail→pass | 6,787 | 2,192 | -68% | 1 | 1 | 0% | 1,254 | 780 | -38% | 0 | 0 | — |
case-08 | fail→pass | 4,016 | 2,396 | -40% | 1 | 1 | 0% | 788 | 835 | +6% | 0 | 0 | — |
case-09 | pass→pass | 5,497 | 4,880 | -11% | 1 | 1 | 0% | 1,096 | 1,456 | +33% | 0 | 0 | — |
case-10 | pass→pass | 4,414 | 4,287 | -3% | 1 | 1 | 0% | 955 | 1,256 | +32% | 0 | 0 | — |
case-11 | pass→pass | 2,358 | 1,990 | -16% | 1 | 1 | 0% | 456 | 798 | +75% | 0 | 0 | — |
case-12 | fail→pass | 4,718 | 2,387 | -49% | 1 | 1 | 0% | 837 | 860 | +3% | 0 | 0 | — |
case-13 | pass→pass | 1,987 | 1,955 | -2% | 1 | 1 | 0% | 348 | 730 | +110% | 0 | 0 | — |
case-14 | pass→pass | 3,712 | 3,840 | +3% | 1 | 1 | 0% | 732 | 1,166 | +59% | 0 | 0 | — |
case-15 | pass→pass | 9,224 | 5,179 | -44% | 1 | 1 | 0% | 1,953 | 1,424 | -27% | 0 | 0 | — |
case-16 | fail→pass | 2,543 | 2,519 | -1% | 1 | 1 | 0% | 498 | 876 | +76% | 0 | 0 | — |
case-17 | fail→pass | 3,406 | 2,900 | -15% | 1 | 1 | 0% | 643 | 906 | +41% | 0 | 0 | — |
case-18 | pass→pass | 10,842 | 6,141 | -43% | 1 | 1 | 0% | 1,860 | 1,513 | -19% | 0 | 0 | — |
case-19 | pass→pass | 5,598 | 3,258 | -42% | 1 | 1 | 0% | 948 | 1,097 | +16% | 0 | 0 | — |
case-20 | fail→pass | 4,477 | 2,847 | -36% | 1 | 1 | 0% | 911 | 954 | +5% | 0 | 0 | — |
case-21 | fail→pass | 3,632 | 2,052 | -44% | 1 | 1 | 0% | 655 | 806 | +23% | 0 | 0 | — |
case-22 | pass→pass | 8,119 | 9,313 | +15% | 1 | 1 | 0% | 1,850 | 2,543 | +37% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +45 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.