Test Writer

SkillFiles & storage

Write tests that catch bugs, with explicit assertions, realistic data, and proper structure. Use when asked to write, add, improve, or expand tests, or to raise coverage. Writes test files and runs them to confirm they assert real behavior.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Test Writer skill

What this skill tells your AI

The instructions your AI receives, as published by thedecipherist/claude-code-mastery-project-starter-kit in .claude/skills/test-writer/SKILL.md and read by ahel’s review.

You write tests that catch bugs, not tests that pass. A test that can't fail isn't a test.

Principles

  1. Every test has explicit assertions. "Page loads" is not a test.
  2. Test behavior, not implementation details.
  3. Cover the happy path, the error cases, and the edge cases.
  4. Use realistic test data, never test / asdf.
  5. Tests are independent. No shared mutable state between them.

Structure

describe('[Feature]', () => {
  describe('[Scenario]', () => {
    it('should [expected behavior] when [condition]', async () => {
      // Arrange — set up test data
      // Act — perform the action
      // Assert — verify SPECIFIC outcomes
    });
  });
});

Assertions

// GOOD — explicit, specific
expect(result.status).toBe(200);
expect(result.body.user.email).toBe('ada@example.com');
await expect(page.locator('h1')).toContainText('Welcome');

// BAD — passes even when broken
expect(result).toBeTruthy();   // too vague
await page.goto('/dashboard'); // no assertion at all

Data-layer tests (this codebase)

The data layer has rules that test data must respect, or the test passes while masking the exact bug that bites in production.

  • Seed real ObjectId values, not string ids. The single most common production bug here is a string-vs-ObjectId _id mismatch that silently returns nothing. A test seeded with string ids passes and hides it. Use actual ObjectId types in fixtures.
  • Exercise the data adapter (StrictDB or native), not a hand-rolled driver mock. Tests go through the same adapters/ boundary the handlers use. Mock at the network or data boundary, not by reimplementing the driver.
  • Test the round trip. Where data is serialized (JSON in, JSON out), assert that types survive it, since that round trip is where _id mismatches and code-66 upsert errors appear.

Unit tests (Vitest)

Each test verifies:

  1. Return value matches expected.
  2. Side effects occurred, or provably didn't.
  3. Error cases throw the proper error.
  4. Edge cases: null, empty, max values, and for ids, wrong-type ids.

E2E tests (Playwright)

Each test verifies:

  1. Correct URL after navigation.
  2. Key elements are present.
  3. Correct data is displayed.
  4. Error states show the proper message.

Before finishing

Run the tests. A new test should fail against code that doesn't satisfy it and pass once it does. If a test passes the moment you write it without the behavior existing, it isn't asserting anything, fix the assertion.

Signals

GitHub stars
338
Forks
40
Last commit
Jun 2026
Advanced
Catalog kind
skill
Gateway key
test-writer-thedecipherist
Source
github.com/thedecipherist/claude-code-mastery-project-starter-kit