Features

This is the complete feature list, grouped by area. For how each check is graded, see Runners; for the Device Lab, see Device Lab.

The runners

Mobile QA runs a fixed catalog of Flutter runners. Only the ones that apply to a project execute; the rest report N/A. Each streams its progress and returns a status, a metric, and a “how to improve” note.

Runner	Category	What it runs	How to use it
Static analysis	Analyze	`flutter analyze --no-pub`	Runs automatically. Fails with the issue count; treat lints as errors in `analysis_options.yaml`.
Unit / widget tests	Tests	`flutter test --coverage --reporter=expanded`	Put tests in `test/`. Fails if any test fails; this also produces the lcov used by Coverage.
Integration tests	Integration	`flutter test integration_test -d <device>`	Add `integration_test/` and pick a device in the run bar. Runs the full app end-to-end.
Golden (snapshot) tests	Golden	`flutter test <golden files>`	Auto-detected from any `matchesGoldenFile` in `test/`. Fails on a pixel diff; update with `flutter test --update-goldens`.
Test coverage	Coverage	reads `coverage/lcov.info`	Run tests with `--coverage` first. Grades line coverage (pass ≥60%).
Test coverage gap	Coverage	maps `lib/` files to tests	Auto-computed. Lists source files with no matching test — the basis of the test plan.
Formatting	Lint	`dart format --set-exit-if-changed --output=none .`	Warns if files need formatting. One-click Auto-fix applies it.
Outdated dependencies	Deps	`flutter pub outdated --no-dev-dependencies`	Informational — lists packages behind their latest.
Startup performance	Performance	`flutter run --profile --trace-startup -d <device>`	Pick a device. Reports time-to-first-frame (aim under ~2s).
Frame / jank profiling	Performance	`flutter test integration_test/perf_test.dart --profile -d <device>`	Scaffold a `perf_test.dart`, pick a device. Reports missed-frame budget %.
App size	Performance	`flutter build apk --analyze-size --target-platform android-arm64 --release`	Builds a release APK and reports total MB + the biggest contributors.
Accessibility	Accessibility	runs `meetsGuideline` tests + statically scans `lib/`	Flags images without `semanticLabel` and icon buttons without `tooltip`.
Release smoke test	Manual QA	— (checklist)	A human sign-off item: install a release build, launch, login, core flow, offline, logout.
Device / OS matrix	Manual QA	— (checklist)	Sign-off: verify on min + latest OS, small + large screen, iOS + Android.
Regression sign-off	Manual QA	— (checklist)	Sign-off: re-test changed areas + previously-fixed bugs before ship.

Grading engine

Every run produces one severity-weighted grade. The score starts at 100 and each failing check deducts by its severity: critical −35, high −20, medium −9, low −3. Warnings and N/A checks don’t deduct. The score maps to a letter — A ≥90, B ≥78, C ≥64, D ≥45, else F — and a per-category Quality summary shows each category’s worst status. See Runners for the full formula.

Auto-fix

One-click, safe-only fixes from the toolbar. Auto-fix runs dart format . then dart fix --apply — formatters only, never your logic — so the change is easy to review. Run it, check the diff, re-run QA.

Per-project trends and sparklines

Mobile QA stores each run per project (the last 20), so it can show how quality is moving:

Score sparkline — a mini line chart of recent scores, on the Dashboard list and the detail header.
Since last run delta — the score change (▲ / ▼), how many checks are new failing, and how many you fixed.
Slack / Telegram alert — if you’ve configured notifications, a run that introduces new failing checks or drops the grade fires an alert automatically.

AI QA

AI features are opt-in — set a provider and key in Settings → AI first (the key stays on your machine and is never sent to the browser). Then, from a project’s detail view:

Test plan — AI drafts a prioritized list of the highest-value test cases to write next, each with type, priority, target file, and given/when/then. Shown as advice; it does not change the grade.
AI test (per file) — the AI button on a test-type card generates a focused widget/unit/integration/golden test for a source file, with a suggested path you can save.
Write scenario — generates BDD/Gherkin scenarios (Feature / Scenario / Given / When / Then) for a file or a free-text feature, then Generate test code turns them into a runnable Flutter test.
Smoke test — writes an integration_test/smoke_test.dart that boots the app and walks its key user journeys end-to-end.
QA summary — a concise, test-readiness briefing from the last run: release-readiness, top risks, and a fix-first order for the week.
Untested files — lists source files with no matching test, each with a one-click AI or Scaffold action.
Explain (AI) — on any failing check, explains the root cause and gives a concrete fix. The Device Lab’s AI triage does the same for a console log.
Fix (AI) — rewrites an offending file to fix the issue, shown as a preview you Apply (keeps a .bak backup).

CI export

From the Reports menu (and the report link) in a project’s detail view:

Report — a standalone, printable HTML QA report with the grade, summary, and every check’s detail + how-to-fix.
Coverage — a per-file coverage table built from coverage/lcov.info, lowest coverage first.
Badge (SVG) — a grade badge for your README or CI, colored by letter.
JUnit XML — one <testcase> per check, for CI dashboards.
JSON — the full run record.
Device matrix — run selected test types across several devices and get a combined report.

The same records are available headless via --qa for CI pipelines (see Usage).

Device Lab

Connect a booted emulator or simulator and check the real app — install & launch, screenshot, screen-record, stream crash-highlighted live logs, and run an automated smoke (Android monkey stress or an iOS integration smoke). Captured screenshots and recordings are saved as artifacts you can preview, download, share, or delete. Full details in Device Lab.

Results and artifacts can go to your team via the shared connectors — Share to team, a Jira issue, and Slack / Telegram / email — configured once in the kit’s connector settings.