Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
different-ai avatar

Testability

  • 12 installs
  • 247 repo stars
  • Updated April 3, 2026
  • different-ai/agent-bank

Makes features testable by design using a testing pyramid from fast unit tests to slow E2E, with dependency injection and test IDs.

About

Teaches designing for testability across unit, API/integration, staging, and E2E layers with patterns like pure-function extraction and mocking only boundaries. A developer uses it before writing code to keep features testable.

  • Testing pyramid mapping speed to layer and when to use each
  • Patterns: dependency injection, extract pure functions, data-testids

Testability by the numbers

  • 12 all-time installs (skills.sh)
  • Ranked #1,527 of 2,153 Testing & QA skills by installs in the Skillselion catalog
  • Data as of Aug 3, 2026 (Skillselion catalog sync)
npx skills add https://github.com/different-ai/agent-bank --skill testability

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Installs12
repo stars247
Last updatedApril 3, 2026
Repositorydifferent-ai/agent-bank

What it does

Makes features testable by design using a testing pyramid from fast unit tests to slow E2E, with dependency injection and test IDs.

Files

SKILL.mdMarkdownGitHub ↗

What I Do

Ensure every feature you build is testable from the start. This skill teaches:

1. Testing pyramid (fast → slow) 2. API exposure patterns for testability 3. Local testing setup 4. Integration with staging tests 5. When to use each testing layer

Core Philosophy

"If you can't test it locally, you can't test it."

Every feature should be testable at multiple levels. Design for testability, don't bolt it on later.

---

Testing Pyramid

From fastest (run constantly) to slowest (run occasionally):

                    ▲
                   /U\        UI Tests (E2E)
                  / I \       - Browser automation
                 /-----\      - Run on staging only
                / API   \     API/Integration Tests
               / TESTS   \    - tRPC procedures
              /-----------\   - Can run locally
             /   UNIT      \  Unit Tests
            /    TESTS      \ - Pure functions
           /------------------\ - Fastest, run always
LayerSpeedWhereWhen to Use
Unit<1sLocalPure logic, utils, calculations
API/Integration1-10sLocal + CItRPC, DB operations, business logic
Staging30s-2mVercel previewFull flow verification
UI/E2E2-5mStaging onlyCritical user journeys

---

Layer 1: Unit Tests (Fastest)

When to Use

  • Pure functions with no side effects
  • Calculations, formatting, validation
  • Business logic that doesn't touch DB/APIs

Pattern

// packages/web/src/lib/utils/calculate-fee.ts
export function calculateFee(amount: number, feePercent: number): number {
  return amount * (feePercent / 100);
}

// packages/web/src/lib/utils/calculate-fee.test.ts
import { describe, it, expect } from 'vitest';
import { calculateFee } from './calculate-fee';

describe('calculateFee', () => {
  it('calculates 1% fee correctly', () => {
    expect(calculateFee(1000, 1)).toBe(10);
  });

  it('handles zero amount', () => {
    expect(calculateFee(0, 5)).toBe(0);
  });
});

Running Unit Tests

cd packages/web
pnpm test                           # Run all tests (watch mode)
pnpm test -- --run                  # Run once and exit
pnpm test:watch                     # Watch mode
pnpm test -- --run --grep "fee"     # Filter by name
Repo note: @zero-finance/web Vitest discovers tests under packages/web/src/test/**/*.test.ts. Put new tests there (or update Vitest config) so they get picked up.

---

Layer 2: API/Integration Tests

When to Use

  • tRPC procedures
  • Database operations
  • External API integrations (mocked)
  • Business logic with dependencies

Pattern: Testing tRPC Procedures

// packages/web/src/server/routers/earn/get-balance.test.ts
import { describe, it, expect, vi, beforeEach } from 'vitest';
import { createTestContext } from '@/test/context';
import { earnRouter } from './index';

describe('earn.getBalance', () => {
  let ctx: ReturnType<typeof createTestContext>;

  beforeEach(() => {
    ctx = createTestContext({
      user: { privyDid: 'test-user-did' },
      workspaceId: 'test-workspace-id',
    });
  });

  it('returns balance for valid user', async () => {
    const caller = earnRouter.createCaller(ctx);
    const result = await caller.getBalance({ chainId: 8453 });

    expect(result).toHaveProperty('balance');
    expect(typeof result.balance).toBe('string');
  });
});

Pattern: Mocking External Services

// Mock Privy
vi.mock('@privy-io/server-auth', () => ({
  PrivyClient: vi.fn().mockImplementation(() => ({
    getUser: vi.fn().mockResolvedValue({ id: 'test-user' }),
  })),
}));

// Mock Database
vi.mock('@/db', () => ({
  db: {
    query: {
      userSafes: {
        findFirst: vi.fn().mockResolvedValue({
          safeAddress: '0x1234...',
          chainId: 8453,
        }),
      },
    },
  },
}));

Test Database Setup

For tests that need a real database:

// packages/web/src/test/setup-db.ts
import { drizzle } from 'drizzle-orm/neon-http';
import { neon } from '@neondatabase/serverless';

export async function createTestDb() {
  // Use a test-specific database
  const sql = neon(process.env.TEST_DATABASE_URL!);
  return drizzle(sql);
}

---

Layer 3: Staging Tests (Vercel Preview)

When to Use

  • Full end-to-end flows
  • Features that involve multiple services
  • UI changes that need visual verification
  • Flows that can't be mocked locally

Workflow

# 1. Push your branch
git push -u origin feat/my-feature

# 2. Wait for deployment
LATEST=$(vercel ls --scope prologe 2>/dev/null | head -1)
vercel inspect "$LATEST" --scope prologe --wait --timeout 5m

# 3. Test on preview URL
# Use Chrome MCP or manual testing

Integration with test-staging-branch Skill

Load the test-staging-branch skill for:

  • Chrome automation login flow
  • Gmail OTP extraction
  • PR reporting
skill("test-staging-branch")

---

Layer 4: UI/E2E Tests (Slowest)

When to Use

  • Critical user journeys only
  • After all other layers pass
  • For regression prevention

Playwright Tests

// packages/web/tests/dashboard.spec.ts
import { test, expect } from '@playwright/test';

test('user can view dashboard balance', async ({ page }) => {
  // Login would use test fixtures
  await page.goto('/dashboard');

  await expect(page.getByText('Total Balance')).toBeVisible();
  await expect(page.getByTestId('balance-amount')).toBeVisible();
});

Running E2E Tests

cd packages/web
pnpm exec playwright test
pnpm exec playwright test --ui  # Interactive mode

---

Making Code Testable

Pattern 1: Dependency Injection

// BAD - Hard to test
export async function getBalance() {
  const safe = await db.query.userSafes.findFirst({...});
  const balance = await fetch(`https://api.example.com/balance/${safe.address}`);
  return balance;
}

// GOOD - Testable
export async function getBalance(
  deps: {
    getSafe: () => Promise<UserSafe>,
    fetchBalance: (address: string) => Promise<string>,
  }
) {
  const safe = await deps.getSafe();
  const balance = await deps.fetchBalance(safe.address);
  return balance;
}

Pattern 2: Extract Pure Functions

// BAD - Logic mixed with I/O
export async function processTransfer(amount: number) {
  const fee = amount * 0.01;
  const total = amount + fee;
  await db.insert(transfers).values({ amount, fee, total });
  return { amount, fee, total };
}

// GOOD - Pure function extractable
export function calculateTransferFees(amount: number) {
  const fee = amount * 0.01;
  const total = amount + fee;
  return { amount, fee, total };
}

export async function processTransfer(amount: number) {
  const calculated = calculateTransferFees(amount);
  await db.insert(transfers).values(calculated);
  return calculated;
}

// Now calculateTransferFees is easily unit testable!

Pattern 3: Test IDs in UI

// Add data-testid for E2E tests
<div data-testid="balance-card">
  <span data-testid="balance-amount">{balance}</span>
</div>

Pattern 4: API Routes for Testing

Expose internal state via API routes that are:

  • Only available in development/test
  • Protected in production
// packages/web/src/app/api/test/state/route.ts
import { NextResponse } from 'next/server';

export async function GET() {
  // Only in development
  if (process.env.NODE_ENV === 'production') {
    return NextResponse.json({ error: 'Not available' }, { status: 403 });
  }

  // Return internal state for testing
  return NextResponse.json({
    safesCount: await db.select().from(userSafes).count(),
    // ... other debug info
  });
}

---

Local Testing Setup

Environment Files

# .env.test - Test-specific config
DATABASE_URL="postgres://test:test@localhost:5432/test_db"
PRIVY_APP_ID="test-app-id"
# ... mocked values

Test Utilities

Create reusable test helpers:

// packages/web/src/test/fixtures.ts
export const testUser = {
  privyDid: 'did:privy:test-user',
  email: 'test@example.com',
};

export const testSafe = {
  address: '0x1234567890123456789012345678901234567890',
  chainId: 8453,
};

// packages/web/src/test/context.ts
export function createTestContext(overrides = {}) {
  return {
    user: testUser,
    workspaceId: 'test-workspace',
    db: mockDb,
    ...overrides,
  };
}

---

Testing Checklist (Per Feature)

Before considering a feature "done":

[ ] Unit tests for pure functions
[ ] Integration tests for tRPC procedures
[ ] Mocks for external services
[ ] Test IDs in UI components
[ ] Manual test on staging (if applicable)
[ ] E2E test for critical paths only

---

Common Anti-Patterns

Don't: Test implementation details

// BAD - Tests internal state
expect(component.state.isLoading).toBe(false);

// GOOD - Tests observable behavior
expect(screen.getByText('Loading...')).not.toBeVisible();

Don't: Over-mock

// BAD - Mock everything
vi.mock('@/db');
vi.mock('@/lib/api');
vi.mock('@/hooks/use-user');
// ... 10 more mocks

// GOOD - Mock only external boundaries
vi.mock('@/lib/external-api'); // Third-party only

Don't: Write E2E tests for everything

// BAD - E2E for simple validation
test('email validation shows error', async ({ page }) => {
  // This should be a unit test!
});

// GOOD - E2E for critical flows only
test('user can complete payment flow', async ({ page }) => {
  // Multi-step, multi-service flow
});

---

Integration with Other Skills

ScenarioSkill to Use
Testing fails on stagingtest-staging-branch
Need to debug prod datadebug prod issues
After completing testsskill-reinforcement
Need Chrome automationchrome-devtools-mcp

---

Learnings Log

Add new learnings here as they're discovered

2024-12-29: Initial skill created

  • Established testing pyramid hierarchy
  • Created patterns for dependency injection and pure function extraction
  • Added integration with other skills

2026-01-12: Next.js 16 async route params

  • Dynamic API routes now receive params as a Promise; await params before reading slug to avoid 404s in local CLI testing.

2026-01-12: Privy user provisioning defaults

  • Privy create-user rejects wallet_index and create_direct_signer; omit defaults and only send those fields when explicitly provided.

---

Quick Reference

# Unit tests
pnpm --filter @zero-finance/web test

# Watch mode
pnpm --filter @zero-finance/web test:watch

# E2E tests
pnpm --filter @zero-finance/web exec playwright test

# Type check (catches many bugs)
pnpm typecheck

# Lint
pnpm lint

Related skills

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.