# Tests

> Test coverage and FIRST — Given/When/Then, business-focused cases, run before ship. Use when adding or changing behavior that needs tests, writing test cases, or verifying coverage before delivery. Don't use for implementing production features without a test obligation, or for unrelated tooling setup.

- Skill: `pedronauck/tests` (Agent Skill)
- Install (CLI): `npx skillmds@latest add pedronauck/tests`
- Raw SKILL.md: https://api.skillmd.com/api/skills/pedronauck/tests/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: pedronauck (https://skillmd.com/u/pedronauck)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/pedronauck/tests

---


# Tests

A change is done only after its tests exist and pass. Apply these rules to every new or altered frontend and backend behavior.

## Steps — ship with tests

**Step 1: Cover the behavior**
1. For every new or altered behavior, add cases for the expected result, errors, boundaries, and missing data.
2. Aim for at least 80% coverage, prioritizing business rules, validation, authorization, error states, and route contracts.
3. Name each case by the observable business outcome; structure it Given/When/Then (or Arrange/Act/Assert) with one business action.

*Done when:* every touched behavior has at least one executed assertion-bearing test, and coverage for the changed surface is ≥80% without assertion-free padding.

**Step 2: Run before delivery**
1. Run the suite and coverage check for the changed project.
2. Fix every failure; re-run the relevant tests after each fix.
3. If no runner or coverage command exists, add one on the first change that requires tests — manual inspection does not replace execution.

*Done when:* the suite and coverage check have been executed after the final fix, coverage is ≥80%, and every failure is resolved.

## Reference — FIRST and focus

- **Fast** — keep unit tests free of network, database, and slow services; reserve integration tests for cross-component contracts.
- **Independent** — each test builds its own state; order and siblings never matter.
- **Repeatable** — isolate uncontrolled externals with mocks, stubs, a frozen clock, and deterministic data.
- **Self-validating** — every relevant result goes through `expect`/`assert`; running code alone is not a test.
- Assert observable rules, input/output contracts, state transitions, and meaningful messages — not private implementation details that can change without changing behavior.