Testing Workflows¶

Claude Code can write tests, run them, fix failures, and help you maintain a healthy test suite. This guide covers strategies for test-driven development, generating test coverage, and keeping tests reliable.

Writing Tests with Claude¶

The simplest approach — ask Claude to write tests for existing code:

bash claude "Write unit tests for src/utils/validation.ts. Cover edge cases like empty strings, special characters, and very long input."

Claude reads the source file, understands the function signatures and logic, then generates tests that cover normal paths, edge cases, and error conditions.

Be Specific About What to Test¶

Vague requests produce vague tests. Tell Claude exactly what concerns you:

```bash

Good: specific coverage goals¶

claude "Write tests for the checkout flow. Make sure to test: applying discount codes, out-of-stock items, and partial refunds."

Less effective: open-ended¶

claude "Write tests for the checkout" ```

Test-Driven Development (TDD)¶

Claude works well in a TDD workflow where you write the test first, then implement the feature:

Step 1: Write the Test¶

bash claude "Write a failing test for a new function called calculateShipping that takes a cart weight in kg and a destination country code, and returns the shipping cost in cents. US orders under 5kg ship for 500 cents, over 5kg for 1200 cents. International is 2500 cents flat."

Step 2: Implement to Pass¶

bash claude "Now implement calculateShipping to make the tests pass."

Step 3: Refactor¶

bash claude "Refactor calculateShipping for clarity. Make sure all tests still pass."

This loop keeps Claude focused and produces well-tested code from the start.

Running Tests¶

Ask Claude to run your test suite directly:

bash claude "Run the tests and fix any failures"

Claude executes your test command (detected from package.json, pytest.ini, Makefile, or your CLAUDE.md), reads the output, and fixes failing tests or the underlying code.

Specify the Test Runner¶

If Claude picks the wrong test command, be explicit:

bash claude "Run tests with: npm run test:unit -- --reporter verbose"

Or add it to your CLAUDE.md:

```markdown

Testing¶

Unit tests: npm run test:unit
Integration tests: npm run test:integration
Run all: npm test ```

Generating Test Coverage¶

Ask Claude to fill gaps in your test suite:

bash claude "Check which functions in src/services/ have no test coverage and write tests for them"

Or target a coverage report:

bash claude "Run the coverage report and write tests to bring src/utils/ above 80% coverage"

Testing Patterns¶

Testing API Endpoints¶

bash claude "Write integration tests for the POST /api/users endpoint. Test successful creation, duplicate email handling, missing required fields, and invalid email format."

Testing with Mocks¶

bash claude "Write tests for the EmailService class. Mock the SMTP client so tests don't send real emails. Verify the correct template is used for each email type."

Snapshot and Approval Tests¶

bash claude "Add snapshot tests for the Dashboard component. If snapshots already exist, update them to reflect the current output."

Testing Error Handling¶

bash claude "Write tests that verify our API returns proper error responses: 400 for bad input, 401 for missing auth, 403 for wrong permissions, 404 for missing resources."

Fixing Flaky Tests¶

Flaky tests are a great use case for Claude:

bash claude "This test fails intermittently: test/integration/websocket.test.ts. Investigate why and fix the flakiness."

Common causes Claude can identify and fix: - Race conditions and missing await calls - Hardcoded ports or timestamps - Test order dependencies - Uncleared state between tests

Maintaining Test Quality¶

Review Generated Tests¶

Always review tests Claude generates. Watch for:

Tests that just mirror the implementation instead of testing behavior
Missing edge cases that you know about from domain knowledge
Overly brittle assertions that break on irrelevant changes

Keep Tests Focused¶

bash claude "This test file has grown to 500 lines. Split it into logical groups and make sure each test file covers one concern."

Update Tests When Code Changes¶

bash claude "I refactored the auth module. Update the existing tests to match the new API while preserving the same coverage."

CI Integration¶

Add test runs to your Claude Code workflow with hooks:

json { "hooks": { "PreToolUse": [ { "matcher": "Bash", "command": "if echo \"$CLAUDE_COMMAND\" | grep -q 'git commit'; then npm test; fi", "description": "Run tests before committing" } ] } }

Or use headless mode in CI pipelines:

bash cat failing-test-output.log | claude -p "Analyze this test failure and suggest a fix"