Flutter TestingCAUTION
A collection of Agent Skills Standard and Best Practice for Programming Languages, Frameworks that help our AI Agent follow best practies on frameworks and programming laguages
Overview
A collection of Agent Skills Standard and Best Practice for Programming Languages, Frameworks that help our AI Agent follow best practies on frameworks and programming laguages
6f5be687a268OBSERVED · 2026-10-14What it tells the agent
The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.
---
name: flutter-testing
description: Write unit, widget, and integration tests with robot patterns, widget keys, and Patrol in Flutter. Use when implementing test behavior; defer CI-only configuration without test changes.
metadata:
triggers:
files:
- '**/test/**.dart'
- '**/integration_test/**.dart'
- '**/robots/**.dart'
- 'lib/core/keys/**.dart'
keywords:
- test
- patrol
- robot
- WidgetKeys
- patrolTest
- blocTest
- mocktail
---
# Flutter Testing Standards
## **Priority: P0 (CRITICAL)**
## The Four-Pillar Test Value Framework
The Four Pillars provide a qualitative reasoning framework to assess test value:
- **Protection against Regressions ($P$)**: Catches real defects; evaluated qualitatively or via mutation kill score. A test that passes when defects exist has zero protection.
- **Resistance to Refactoring ($R$)**: Decoupled from implementation details; zero false alarms on internal refactorings or domain additions. Tests coupled to private state or mock sequences score low on resistance.
- **Fast Feedback ($F$)**: Executes in milliseconds; rapid local TDD feedback loop.
- **Maintainability ($M$)**: Clean AAA structure, fluent test data builders when setup repetition obscures intent, high readability, zero boilerplate duplication.
Avoid calculating or inventing arbitrary numeric scores (e.g. $P \times R \times F \times M$) statically without measured mutation or runtime evidence.
## Core Rule Anchors
- **`[MOB-TEST-01]` Tripartite Naming**: Name tests `Method_Scenario_ExpectedBehavior`.
- **`[MOB-TEST-02]` BLoC State Invariant Rule**: Assert state transitions using `isA<State>().having(...)` predicates; BAN full copyWith mirrors.
- **`[MOB-TEST-03]` Entity Invariant & Serialization Rule**: Test calculations, validations, domain invariants, and non-trivial serialization/parsing or error mapping. BAN testing generated code, trivial getters/setters, `props`, or echo tests (`expect(item.discount, equals(5))`) where the test merely repeats literal assignments.
- **`[MOB-TEST-04]` Contract Testing Rule**: Repositories and data sources must be tested for contract compliance, golden JSON parsing, and error mapping. BAN 1:1 pass-through mock echoing (`when(() => dataSource.get()).thenAnswer((_) async => x); expect(await repo.get(), x)`).
- **`[MOB-TEST-05]` Bug-First Regression Lock**: Every PR fixing a bug ticket or with title `fix(...)` must introduce a failing test reproducing the defect before fixing it.
## Core Rules
1. **Test Pyramid**: Unit > Widget > Integration.
2. **Naming**: Follow `[MOB-TEST-01]` (`Method_Scenario_ExpectedBehavior`).
3. **AAA**: Arrange, Act, Assert in all tests.
4. **Isolated Builders**: Fluent builders when setup complexity warrants (`OrderBuilder`); avoid brittle shared global mocks.
5. **File Placement**: `_integration_test.dart` ONLY in `integration_test/`.
6. **Robot-First**: All UI assertions/interactions via **Robot pattern** (extending `BaseRobot`) — never raw `find.*`/`expect()` in test body. Use `WidgetKeys` constants from `lib/core/keys/`.
7. **Negative Assertions**: Add `expectXxxNotVisible()` in robot only when absence is part of the business contract (e.g. restricted action hidden), not blanket pairs for unrelated content.
8. **Widget Testing & Mocking**: Setup with `TestWrapper.init()` and `tester.pumpLocalizedWidget(...)`. Register mock BLoCs with `GetIt` in `setUpAll`; stub `state` and `stream` in `setUp` (use `whenListen` and `settle: false` for loading/transition states). Prohibit `any()`.
9. **Integration Testing**: Use `patrolTest` with `IntegrationAuthHelper.loginOrSkip($)` for auth flows and `native interactions` (`$.native.*`).
## Anti-Patterns & Banned Smells
- **`[MOB-TEST-01]` Vague names**: Banish non-descriptive names (`testCart`).
- **`[MOB-TEST-02]` Brittle state mirrors**: Ban 30-property `copyWith` trees in `blocTest`; use `isA<State>().having(...)`.
- **`[MOB-TEST-03]` Entity echo tests**: Ban asserting trivial field assignments or auto-generated boilerplate (`discount: 5`). Meaningful serialization/parsing error handling remains valid.
- **`[MOB-TEST-04]` Pass-through mock echoing**: Ban 1:1 repository-to-datasource pass-through mocks without contract assertions.
- **`[MOB-TEST-05]` Missing bug reproduction**: Ban bug fix PRs without reproduction tests.
- **No blanket negative assertions**: Avoid asserting absence of unrelated content.
- **No inline Key**: Use `WidgetKeys` constant. **No `any()`**: Use typed matchers.
- **No local mocks**: Use `test/shared/`. **No missing bloc stub**: Stub `state` + `stream`.
- **No test-body logic**: Move `find.*`/`expect()` to robot. No raw find in integration tests.
- **No unchecked text casing**: Verify `.toUpperCase()`, `.tr()` in source.
## Verification
- [ ] Repositories use fakes over mocks where appropriate.
- [ ] Distinct BLoC/ViewModel behaviors covered (loading, success, error) with `isA<State>().having(...)`.
- [ ] Views with distinct UI logic tested via Robot pattern at proper layer.
- [ ] Critical user flows have at least one integration test using `patrolTest`.
- [ ] `flutter test` passes.
## Canonical response anchors
When this skill applies, preserve the following domain terminology or equivalent concrete examples in the answer when relevant:
- GetIt registration
- Register the mock with `GetIt` in `setUpAll` before building the widget under test.
- native interactionsTrust audit
CAUTIONgrade C · trust 79/100 Install with care. The audit found things worth knowing before you trust its output.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | NA |
| L2 | Instruction surface (what it tells the agent) | FAIL |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- none-observed
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (3)
Prohibit `any()` / `anyNamed()` — bypass type safety, cause silent failures. Use specific values or typed matchers.
await robot.loginWith(email: '[email protected]', password: 'pass');
await robot.loginWith(email: '[email protected]', password: 'pass');
Gates applied: instruction_override, no_behavioural_pass.
6f5be687a268full audit observations/trust-audit/skill/hoangnguyen0403__flutter-testing.json · Report an issue / request a re-scanAudit history
Every audit this skill has had.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-14 | 6f5be687a268 | CAUTION | C | 79 | first audit |
Questions
What does the Flutter Testing skill do?
A collection of Agent Skills Standard and Best Practice for Programming Languages, Frameworks that help our AI Agent follow best practies on frameworks and programming laguages
Is Flutter Testing safe to install?
With care. The audit graded it C (79/100) and found 3 things worth knowing before you trust this skill, listed below with the exact line each was found on.
What can Flutter Testing access on my machine?
The audit observed no filesystem, network or shell use at all in its source.
How current is this page?
The grade is for one exact copy of the source (6f5be687a268), read on 2026-10-14. The repository is watched, and a new audit runs when it changes — this is the first audit.