test-driven-development

name	test-driven-development
description	Drive implementation with tests. Use when fixing bugs, changing behavior, adding logic, modifying contracts, or proving that agent-written code works and remains safe across regressions.

Test-Driven Development

Skill Interface

• Name: test-driven-development.
• Description: Drive implementation with tests when fixing bugs, changing behavior, adding logic, modifying contracts, or proving that agent-written code works and remains safe across regressions.
• Parameters: Desired behavior, bug reproduction, affected logic or contract, existing tests, suitable test layer, fixtures or inputs, and available test commands.
• Instructions: Use this skill before or during behavior changes. Start with a failing or relevant test, make the smallest implementation change, refactor while tests stay green, and run adjacent regression checks.

Use tests to prove behavior, not to document implementation trivia.

Cycle

1. Red: write or identify a failing test that captures the required behavior or bug.
2. Green: make the smallest implementation change that passes.
3. Refactor: simplify while keeping tests green.
4. Regression: run adjacent existing tests that could be affected.

Bug Fix Pattern

1. Reproduce the bug with a failing test, fixture, command, trace, or fixed input/output pair.
2. Confirm the failure matches the reported issue.
3. Fix the root cause.
4. Re-run the same reproduction and relevant regression tests.
5. Keep the regression test unless it is too expensive and a documented manual check is more appropriate.

Test Selection

• Unit tests for deterministic logic.
• Integration tests for boundaries, adapters, persistence, and schemas.
• Contract tests for APIs, events, messages, and error envelopes.
• End-to-end tests for critical user workflows.
• Smoke tests for live providers only when environment and cost are acceptable.

Rules

• Do not delete failing tests to get a green run.
• Do not loosen assertions unless the old assertion was wrong.
• Do not mock the core behavior being tested.
• Do not hide races with fixed sleeps.
• Do not claim unexecuted tests passed.

Coverage Checklist

Consider:

• Normal success.
• Invalid input.
• Unknown fields or variants.
• Upstream failure.
• Permission denied.
• Timeout and cancellation.
• Retry or duplicate event.
• Partial result or degraded behavior.