AgentSkillsCN

test-engineer

为 TypeScript/React(Vitest)和 Rust(cargo test)代码编写有意义的测试用例。

SKILL.md
--- frontmatter
name: test-engineer
description: Write meaningful tests for TypeScript/React (Vitest) and Rust (cargo test) code.

Test Engineer Skill

Write tests that fail when implementation breaks. No placeholders, no expect(true).toBe(true).

Process

Step 1: Read the File

Use Read tool to read the entire file. Identify:

  • Type: React component, Zustand store, Zod schema, Rust module, Tauri command
  • Public API: Exported functions, actions, props, methods
  • Dependencies: Imports, external calls (Tauri invoke, APIs)
  • Middleware: persist, immer, devtools (for Zustand stores)
  • Side effects: API calls, state mutations, filesystem operations

Step 2: Output Analysis (MANDATORY)

Before writing ANY test code, output:

code
## Analysis: [filename]

**Type:** [React Component | Zustand Store | Zod Schema | Rust Module | Tauri Command]
**Purpose:** [1-2 sentences]
**Middleware:** [persist | immer | none] (if Zustand)

**Public API:**
- `functionName`: [what it does]

**Test Scenarios:**
1. [Happy path]
2. [Edge case]
3. [Error case]

**Skip Testing:**
- [Any pass-through functions or trivial code]

Step 3: Write Tests

See references/templates.md for framework-specific templates.

Framework Selection

File TypeFrameworkMock Strategy
Zustand Store (/stores/*.ts)VitestMock @tauri-apps/api/core if store calls invoke
Zustand + persistVitestTest partialize and merge callbacks
React Component (*.tsx)Vitest + RTLCustom render with providers
Zod Schema (/schemas/*.ts)VitestNone needed
Rust Module (*.rs)cargo testtempfile for filesystem
Tauri Command (/commands/*.rs)#[tokio::test]tempfile, real backend

Critical Rules

1. Every Assertion Must Be Meaningful

Before writing any assertion, ask: "If I delete the implementation, will this fail?"

  • No → Delete the test or rewrite it
  • Yes → Keep it

2. What NOT to Test

Skip these entirely:

  • Pure re-exports or pass-through functions
  • Type definitions and interfaces
  • Constants (unless they affect behavior)
  • Auto-generated code
  • Private/internal functions (test through public API)
  • Logging statements (don't test logger.debug was called)
  • Simple getters/setters with no logic
  • Selector hooks that just re-export store slices (e.g., useInputMode = () => useStore(s => s.inputMode))
  • Third-party library wrappers that add no logic

3. Test Behaviors, Not Implementation

typescript
// BAD: Testing internal state shape
expect(store.getState()._internal.cache).toHaveLength(3);

// GOOD: Testing observable behavior
expect(store.getState().items).toHaveLength(3);

4. Coverage Requirements

For each public function, test:

  • Happy path: Valid input → expected output
  • Edge cases: Empty, null, undefined, boundaries, deduplication
  • Error cases: Invalid input, failures, exceptions
  • Async (if applicable): Loading states, error states, race conditions

For Zustand stores with middleware:

  • persist: Test partialize returns correct subset, merge handles corruption
  • Computed getters: Test functions like getContextPercentage() with known inputs

5. Test Independence

  • Reset state in beforeEach
  • No shared mutable state between tests
  • Each test must pass when run alone

Zustand Reset Pattern:

typescript
beforeEach(() => {
  // Prefer reset() if the store exports it
  const store = useStore;
  const state = store.getState();
  if ('reset' in state && typeof state.reset === 'function') {
    state.reset();
  } else if (typeof store.getInitialState === 'function') {
    // Zustand v5: restores full initial state including actions
    store.setState(store.getInitialState(), true);
  } else {
    // Last resort: reset known fields WITHOUT replace (avoid dropping actions)
    store.setState({
      items: [],
      isLoading: false,
      error: null,
    });
  }
  vi.clearAllMocks();
});

Important: Do NOT use setState(partial, true) unless you're passing the full initial state including actions. Replacing with a partial object will drop actions and break the store.

Immer + Set/Map: If the store uses Set or Map with immer middleware, add at top of test file:

typescript
import { enableMapSet } from 'immer';
enableMapSet();

Anti-Patterns (Never Do These)

See references/anti-patterns.md for detailed examples.

typescript
// NEVER: Always-passing assertion
it('should work', () => {
  const store = useStore.getState();
  expect(store).toBeDefined(); // Passes even if store is broken
});

// NEVER: Testing implementation details
it('should set internal flag', () => {
  action();
  expect(store.getState()._hasLoaded).toBe(true);
});

// NEVER: Placeholder test
it('should handle edge cases', () => {
  // TODO: implement
});

Tauri-Specific Patterns

See references/tauri-mocks.md for composable mock patterns.

Key principle: Only mock Tauri if the code under test calls invoke directly. Many Zustand stores don't.

Composable invoke mock (accumulates, doesn't overwrite):

typescript
import { vi } from 'vitest';
import { invoke } from '@tauri-apps/api/core';

vi.mock('@tauri-apps/api/core', () => ({ invoke: vi.fn() }));

// In test setup file - creates composable mock
const commandMocks = new Map<string, unknown>();

export function mockCommand(cmd: string, response: unknown | Error) {
  commandMocks.set(cmd, response);
}

export function clearCommandMocks() {
  commandMocks.clear();
}

// Set up once globally
vi.mocked(invoke).mockImplementation(async (cmd: string) => {
  if (commandMocks.has(cmd)) {
    const response = commandMocks.get(cmd);
    if (response instanceof Error) throw response;
    return response;
  }
  throw new Error(`Unmocked command: ${cmd}`);
});

React Testing Library Query Guide

Use the right query for the situation:

Query TypeWhen to UseThrows if Missing?
getBy*Element should exist NOWYes
findBy*Element will appear ASYNCYes (after timeout)
queryBy*Element might NOT existNo (returns null)
typescript
// Element exists synchronously
expect(screen.getByRole('button')).toBeInTheDocument();

// Element appears after async operation
await waitFor(() => {
  expect(screen.getByText('Loaded')).toBeInTheDocument();
});
// OR
const button = await screen.findByRole('button', { name: 'Submit' });

// Assert element does NOT exist
expect(screen.queryByText('Error')).not.toBeInTheDocument();

Custom Render with Providers

Most components need providers. Create a custom render:

typescript
// src/test/utils.tsx
import { render, type RenderOptions } from '@testing-library/react'
import { ThemeProvider } from '@/providers/theme-provider'

const AllProviders = ({ children }: { children: React.ReactNode }) => (
  <ThemeProvider>
    {children}
  </ThemeProvider>
)

const customRender = (
  ui: React.ReactElement,
  options?: Omit<RenderOptions, 'wrapper'>
) => render(ui, { wrapper: AllProviders, ...options })

export * from '@testing-library/react'
export { customRender as render }

Self-Check Before Finishing

  • Every test has meaningful assertion that tests real behavior
  • No expect(true).toBe(true) or expect(x).toBeDefined() without follow-up
  • No placeholder/TODO tests
  • All public API covered
  • Edge cases tested (empty, null, boundaries, deduplication)
  • Error cases tested
  • State reset in beforeEach (using reset() or manual setState)
  • Tests independent (no shared mutable state)
  • Middleware tested if applicable (persist partialize/merge)
  • Computed getters tested if they have logic

Running Tests

Frontend Apps (Vitest):

bash
bun run test                                 # Run all frontend tests
bun run test apps/agent/src/path/to/test.ts  # Run specific file
bun run test:watch                           # Watch mode
bun run test:coverage                        # With coverage report

Note: Uses Vitest with jsdom for DOM/React testing. Config: vitest.config.ts

agent-bridge (Bun Test):

bash
cd agent-bridge && bun test                  # Bun runtime tests (no jsdom)

Rust Backend (Cargo):

bash
cargo test                                   # Run all Rust tests
cargo test test_function_name                # Run specific test
cargo test -- --nocapture                    # Show println! output