# TDD Enforcement

> TDD Enforcement is an Agent Skill for testing & quality published in the a5c-ai/babysitter repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Red-Green-Refactor TDD methodology with mandatory failing tests, minimal implementation, quality refactoring, and 80% coverage gating.

- Skill name: `tdd-enforcement`
- Category: [Testing & Quality](https://www.openparable.com/skill/testing-quality/1)
- Source: https://github.com/a5c-ai/babysitter/blob/HEAD/library/methodologies/everything-claude-code/skills/tdd-enforcement/SKILL.md
- Repository: a5c-ai/babysitter (1,548 GitHub stars when collected)

## SKILL.md

- Write tests that define expected behavior
- Tests MUST fail (exit code 1)
- Use CI=true or --run flag, never watch mode
- Apply timeout guards (60s) to prevent hanging
- Record exit code as evidence

### 2. GREEN Phase - Minimal Implementation
- Write the minimal code to make tests pass
- Do NOT add features not covered by tests
- Do NOT optimize prematurely
- Tests MUST pass (exit code 0)
- Record exit code as evidence

### 3. REFACTOR Phase - Quality Improvement
- Apply SOLID principles and clean code patterns
- Improve naming, reduce coupling
- Remove duplication
- Run tests after EACH refactoring step
- Tests MUST remain passing (exit code 0)

### 4. Coverage Gate
- Measure coverage: statements, branches, functions, lines
- Minimum 80% overall coverage required
- Iterate: write additional tests for gaps until threshold met
- Maximum 3 convergence iterations

## Rules

- Never skip the RED phase
- Never accept GREEN without exit code 0
- Never use watch mode in CI
- Always record evidence (exit codes, coverage numbers)
- Enforce 80% coverage threshold

## When to Use

- All code implementation tasks
- Feature development
- Bug fixes (write regression test first)

## Agents Used

- `tdd-guide` (primary consumer)
- `code-reviewer` (validates test quality)

---

Source: https://www.openparable.com/skill/testing-quality/tdd-enforcement (OpenParable). Cite the canonical URL above.
