Test-Driven Development
TDD is the fundamental practice. Every line of production code must be written in response to a failing test.
For how to write good tests, load the testing skill. This skill focuses on the TDD workflow/process.
RED-GREEN-REFACTOR Cycle
RED: Write Failing Test First
-
NO production code until you have a failing test
-
Test describes desired behavior, not implementation
-
Test should fail for the right reason
GREEN: Minimum Code to Pass
-
Write ONLY enough code to make the test pass
-
Resist adding functionality not demanded by a test
-
Commit immediately after green
REFACTOR: Assess Improvements
-
Assess AFTER every green (but only refactor if it adds value)
-
Commit before refactoring
-
All tests must pass after refactoring
TDD Evidence in Commit History
Default Expectation
Commit history should show clear RED → GREEN → REFACTOR progression.
Ideal progression:
commit abc123: test: add failing test for user authentication commit def456: feat: implement user authentication to pass test commit ghi789: refactor: extract validation logic for clarity
Rare Exceptions
TDD evidence may not be linearly visible in commits in these cases:
- Multi-Session Work
-
Feature spans multiple development sessions
-
Work done with TDD in each session
-
Commits organized for PR clarity rather than strict TDD phases
-
Evidence: Tests exist, all passing, implementation matches test requirements
- Context Continuation
-
Resuming from previous work
-
Original RED phase done in previous session/commit
-
Current work continues from that point
-
Evidence: Reference to RED commit in PR description
- Refactoring Commits
-
Large refactors after GREEN
-
Multiple small refactors combined into single commit
-
All tests remained green throughout
-
Evidence: Commit message notes "refactor only, no behavior change"
Documenting Exceptions in PRs
When exception applies, document in PR description:
TDD Evidence
RED phase: commit c925187 (added failing tests for shopping cart) GREEN phase: commits 5e0055b, 9a246d0 (implementation + bug fixes) REFACTOR: commit 11dbd1a (test isolation improvements)
Test Evidence: ✅ 4/4 tests passing (7.7s with 4 workers)
Important: Exception is for EVIDENCE presentation, not TDD practice. TDD process must still be followed - these are cases where commit history doesn't perfectly reflect the process that was actually followed.
Coverage Verification - CRITICAL
NEVER Trust Coverage Claims Without Verification
Always run coverage yourself before approving PRs.
Verification Process
Before approving any PR claiming "100% coverage":
Check out the branch
git checkout feature-branch
Run coverage verification:
cd packages/core pnpm test:coverage
OR
pnpm exec vitest run --coverage
Verify ALL metrics hit 100%:
-
Lines: 100% ✅
-
Statements: 100% ✅
-
Branches: 100% ✅
-
Functions: 100% ✅
Check that tests are behavior-driven (not testing implementation details)
For anti-patterns that create fake coverage (coverage theater), see the testing skill.
Reading Coverage Output
Look for the "All files" line in coverage summary:
| File | % Stmts | % Branch | % Funcs | % Lines | Uncovered Line #s |
|---|---|---|---|---|---|
| All files | 100 | 100 | 100 | 100 | |
| setup.ts | 100 | 100 | 100 | 100 | |
| context.ts | 100 | 100 | 100 | 100 | |
| endpoints.ts | 100 | 100 | 100 | 100 |
✅ This is 100% coverage - all four metrics at 100%.
Red Flags
Watch for these signs of incomplete coverage:
❌ PR claims "100% coverage" but you haven't verified
- Never trust claims without running coverage yourself
❌ Coverage summary shows <100% on any metric
All files | 97.11 | 93.97 | 81.81 | 97.11 |
- This is NOT 100% coverage (Functions: 81.81%, Lines: 97.11%)
❌ "Uncovered Line #s" column shows line numbers
setup.ts | 95.23 | 100 | 60 | 95.23 | 45-48, 52-55
- Lines 45-48 and 52-55 are not covered
❌ Coverage gaps without explicit exception documentation
- If coverage <100%, exception should be documented (see Exception Process below)
When Coverage Drops, Ask
"What business behavior am I not testing?"
NOT "What line am I missing?"
Add tests for behavior, and coverage follows naturally.
100% Coverage Exception Process
Default Rule: 100% Coverage Required
No exceptions without explicit approval and documentation.
Requesting an Exception
If 100% coverage cannot be achieved:
Step 1: Document in package README
Explain:
-
Current coverage metrics
-
WHY 100% cannot be achieved in this package
-
WHERE the missing coverage will come from (integration tests, E2E, etc.)
Step 2: Get explicit approval
From project maintainer or team lead
Step 3: Document in CLAUDE.md
Under "Test Coverage: 100% Required" section, list the exception
Example Exception:
Current Exceptions
- Next.js Adapter: 86% function coverage
- Documented in
/packages/nextjs-adapter/README.md - Missing coverage from SSR functions (tested in E2E layer)
- Approved: 2024-11-15
- Documented in
Remember
The burden of proof is on the requester. 100% is the default expectation.
Development Workflow
Adding a New Feature
-
Write failing test - describe expected behavior
-
Run test - confirm it fails (pnpm test:watch )
-
Implement minimum - just enough to pass
-
Run test - confirm it passes
-
Refactor if valuable - improve code structure
-
Commit - with conventional commit message
Workflow Example
1. Write failing test
it('should reject empty user names', () => { const result = createUser({ id: 'user-123', name: '' }); expect(result.success).toBe(false); }); # ❌ Test fails (no implementation)
2. Implement minimum code
if (user.name === '') { return { success: false, error: 'Name required' }; } # ✅ Test passes
3. Refactor if needed (extract validation, improve naming)
4. Commit
git add . git commit -m "feat: reject empty user names"
Commit Messages
Use conventional commits format:
feat: add user role-based permissions fix: correct email validation regex refactor: extract user validation logic test: add edge cases for permission checks docs: update architecture documentation
Format:
-
feat:
-
New feature
-
fix:
-
Bug fix
-
refactor:
-
Code change that neither fixes bug nor adds feature
-
test:
-
Adding or updating tests
-
docs:
-
Documentation changes
Pull Request Requirements
Before submitting PR:
-
All tests must pass
-
All linting and type checks must pass
-
Coverage verification REQUIRED - claims must be verified before review/approval
-
PRs focused on single feature or fix
-
Include behavior description (not implementation details)
Example PR Description:
Summary
Adds support for user role-based permissions with configurable access levels.
Behavior Changes
- Users can now have multiple roles with fine-grained permissions
- Permission check via
hasPermission(user, resource, action) - Default role assigned if not specified
Test Evidence
✅ 42/42 tests passing ✅ 100% coverage verified (see coverage report)
TDD Evidence
RED: commit 4a3b2c1 (failing tests for permission system) GREEN: commit 5d4e3f2 (implementation) REFACTOR: commit 6e5f4a3 (extract permission resolution logic)
Refactoring Priority
After green, classify any issues:
Priority Action Examples
Critical Fix now Mutations, knowledge duplication, >3 levels nesting
High This session Magic numbers, unclear names, >30 line functions
Nice Later Minor naming, single-use helpers
Skip Don't change Already clean code
For detailed refactoring methodology, load the refactoring skill.
Anti-Patterns to Avoid
-
❌ Writing production code without failing test
-
❌ Testing implementation details (spies on internal methods)
-
❌ 1:1 mapping between test files and implementation files
-
❌ Using let /beforeEach for test data
-
❌ Trusting coverage claims without verification
-
❌ Mocking the function being tested
-
❌ Redefining schemas in test files
-
❌ Factories returning partial/incomplete objects
-
❌ Speculative code ("just in case" logic without tests)
For detailed testing anti-patterns, load the testing skill.
Summary Checklist
Before marking work complete:
-
Every production code line has a failing test that demanded it
-
Commit history shows TDD evidence (or documented exception)
-
All tests pass
-
Coverage verified at 100% (or exception documented)
-
Test factories used (no let /beforeEach )
-
Tests verify behavior (not implementation details)
-
Refactoring assessed and applied if valuable
-
Conventional commit messages used