Imported from Ahnd6474/Jakal-flow (.skill-staging/skills/accelint-ts-testing/AGENTS.md). Install upstream with npx skills add Ahnd6474/Jakal-flow --skill accelint-ts-testing. Copyright stays with the author.

Vitest Best Practices

Note: This document is mainly for agents and LLMs to follow when maintaining, generating, or refactoring vitest tests. Humans may also find it useful, but guidance here is optimized for automation and consistency by AI-assisted workflows.

Abstract

Expert-level vitest testing guidance designed for AI agents and LLMs. Each rule includes one-line summaries here, with links to detailed examples in the references/ folder. Load reference files only when you need detailed implementation guidance for a specific rule.

Token efficiency principle: This guide focuses on expert-level insights and non-obvious patterns. It assumes understanding of basic vitest concepts (describe, it, expect, vi) and focuses on decisions experts make: when to mock vs use real code, how to structure tests for maintainability, and performance optimization patterns.

How to Use This Guide

1. Start here: Scan the rule summaries to identify relevant patterns
2. Check for existing setup files: Before writing tests, look for setup files that configure global mocks and utilities
3. Load references as needed: Click through to detailed examples only when implementing
4. Progressive loading: Each reference file is self-contained with examples

This structure minimizes context usage while providing complete implementation guidance when needed.

Workflow: Before Writing Tests

0. Verify the file contains testable behavior Before writing any tests, check if the file actually needs testing:

• Does it contain functions or logic? → Test it
• Does it only export constants, types, or data? → Skip testing it

Files without behavior (constants files, type definitions, GLSL uniform declarations, pure data files) don't need tests. Testing expect(CONSTANT).toBe(value) provides no value and wastes CI time.

1. Check vitest.config.ts for global configuration Verify mock cleanup is configured globally:

// Look for these settings:
clearMocks: true      // Mock cleanup configured?
mockReset: true       // Mock reset configured?
restoreMocks: true    // Mock restore configured?

If not present, recommend adding them. This eliminates the entire class of mock cleanup errors.

2. Discover existing test setup files Check common locations for test setup configuration:

• test/setup.{ts,js} or testing/setup.{ts,js}
• vitest.setup.{ts,js} or src/test/setup.{ts,js}
• Check vitest.config.ts for configured setupFiles and globalSetup

3. Analyze setup file contents When found, identify:

• Global mocks (fetch, timers, etc.)
• Custom matchers (e.g., @testing-library/jest-dom)
• Test utilities and helpers
• Environment configuration

4. Only add per-test cleanup for non-mock resources If global config handles mocks, DO NOT add manual mock cleanup:

• ❌ Don't add vi.clearAllMocks() (handled by config)
• ✅ Do clean up listeners, connections, custom state

Principle: Configuration over repetition Mock cleanup is a safety concern. Configure it once globally to make forgetting impossible. Manual cleanup in every test violates DRY and creates maintenance burden.

See vitest-features.md and performance.md for detailed examples.

Workflow: Before Marking Test Files Complete

CRITICAL: This workflow is MANDATORY. Never skip type checking test files.

Before marking any test file as "complete" or "done", verify type correctness:

Why this matters: Test files are typically excluded from tsconfig.json compilation (not in include paths), so running tsc at the project root won't catch type errors in tests. Type errors in tests can cause:

• Runtime failures that should have been caught at compile time
• Incorrect test behavior due to type mismatches
• False confidence from tests that don't actually test what they claim

Verification steps:

1. Navigate to the package directory (CRITICAL for monorepos): For monorepos or multi-package projects, you MUST cd into the specific package directory before running type checking. TypeScript needs to run from where the tsconfig.json and node_modules are located for that package.

# Example for monorepo:
cd packages/my-package
# Then run tsc from here

2. Check test file directly with TypeScript: Use the project's package manager to run TypeScript:

# Detect which package manager to use:
# - npm: npm exec tsc -- --noEmit path/to/test.test.ts
# - pnpm: pnpm exec tsc --noEmit path/to/test.test.ts
# - bun: bunx tsc --noEmit path/to/test.test.ts
# - yarn: yarn exec tsc --noEmit path/to/test.test.ts

To detect the package manager, check for:

• bun.lockb or bun.lock → use bunx
• pnpm-lock.yaml → use pnpm exec
• yarn.lock → use yarn exec
• package-lock.json → use npm exec

2. Look for common type issues:

• Mock types not matching actual implementation types
• Test data with missing or incorrect properties
• Assertion types that don't match expected values
• Missing type parameters on generic functions
• Incorrect use of type guards or type assertions

3. Fix all type errors before marking complete (NON-NEGOTIABLE) This step is MANDATORY, not optional. Type errors in tests are as critical as type errors in production code.

• Do NOT use as any or @ts-ignore to bypass type checking
• Update test data to match actual types
• Fix mock return types to match implementation
• Add proper type annotations where TypeScript cannot infer
• If you encounter type errors, STOP and fix them - do not proceed with "I'll fix types later"

Example type errors to catch:

// ❌ Type error: property 'email' is missing
const user = createUser({ name: 'Alice' })

// ✅ Correct: all required properties provided
const user = createUser({ name: 'Alice', email: 'alice@example.com' })

// ❌ Type error: vi.fn() returns unknown, not User
const mockGetUser = vi.fn().mockReturnValue({ id: 1 })

// ✅ Correct: explicitly type the mock
const mockGetUser = vi.fn<() => User>().mockReturnValue({ id: 1, name: 'Alice', email: 'test@example.com' })

Principle: Type-safe tests prevent silent failures Type errors in tests are as critical as type errors in production code. Catch them before marking work complete.

CRITICAL REMINDER: This workflow is NOT optional. Running tsc --noEmit against test files is a REQUIRED step before marking test work as complete. If you skip this step, you risk shipping broken tests that provide false confidence.

Workflow: Test Code Review/Audit

When reviewing existing test code (skill invoked with file path or user asks to "review tests" or "audit tests"), follow this systematic approach:

1. Load property-based-testing.md for pattern detection Always load property-based-testing.md during test audits to check for PBT opportunities.

2. Identify anti-patterns and violations Check for violations of rules in sections 1.1-1.10 below.

3. Check for property-based testing opportunities For each test file, analyze the code under test and identify high-value PBT patterns:

ALWAYS check for these patterns:

• Encode/decode pairs: Functions like encode()/decode(), serialize()/deserialize(), toJSON()/fromJSON() → Suggest roundtrip property
• Normalization functions: normalize(), sanitize(), format() → Suggest idempotence property
• Validator + normalizer pairs: isValid() + normalize() → Suggest "isValid(normalize(x)) always true"
• Pure transformation functions: No side effects, deterministic → Multiple properties may apply
• Sorting/ordering functions: sort(), compare() → Suggest ordering + idempotence properties
• Data structure operations: Custom collections with invariants → Suggest invariant properties

When identifying PBT opportunities:

• Check if fast-check is installed (package.json devDependencies)
• If installed: Recommend PBT improvements directly
• If NOT installed: Suggest PBT as an option with user approval required

4. Generate report using template Use assets/output-report-template.md and include PBT opportunities in a dedicated section.

Example PBT opportunity detection:

// Code under test:
function get<T>(obj: Record<string, T>, path: string): T | undefined

// Example-based test found:
it('gets nested value', () => {
  expect(get({ a: { b: 1 } }, 'a.b')).toBe(1)
})

// PBT opportunity identified:
// ✅ EXCELLENT CANDIDATE for property-based testing
// Pattern: Pure function with clear invariants
// Properties to test:
// 1. get(obj, path) returns undefined for non-existent paths
// 2. get(obj, path) preserves type (type preservation)
// 3. get(obj, path) never throws on valid inputs

Principle: Proactive improvement suggestions Don't wait for users to ask "would any benefit from PBT?" — proactively identify and suggest PBT opportunities as part of every test audit.

1. General

1.1 Organization

Place test files next to implementation; one test file per module. View detailed examples

1.2 AAA Pattern

Structure tests as Arrange, Act, Assert for clarity. View detailed examples

1.3 Parameterized Tests

Use it.each for variations; one behavior per test. View detailed examples

1.4 Error Handling

Test negative cases, fault injection, and recovery thoroughly. View detailed examples

1.5 Assertions

Use strict assertions (toEqual, toStrictEqual) over loose ones. View detailed examples

1.6 Test Doubles

Prefer fakes > stubs > spies/mocks; avoid over-mocking. View detailed examples

1.7 Async Testing

Test promises, async/await, and timers correctly. View detailed examples

1.8 Performance

Keep tests fast through efficient setup and avoiding expensive operations. View detailed examples

1.9 Vitest Features

Use coverage, watch mode, benchmarking, and other vitest-specific features. View detailed examples

1.10 Snapshot Testing

Use snapshots for appropriate cases; avoid common pitfalls. View detailed examples

1.11 Property-Based Testing

Use fast-check for stronger coverage with generated inputs; test encode/decode pairs, validators, normalizers, and invariants. View detailed examples