Add tdd-loop skill — autonomous TDD loop for Claude Code and Claude.ai

[Jdevgame]

Add tdd-loop skill

What it does

Enforces a strict Test-Driven Development loop automatically — no need to
remind Claude to write tests first. Every time you ask Claude to implement,
fix, or change code, it:

1. Runs a confidence check before writing any test (unique mechanism —
   Claude scores its own understanding 0–100% and gates on your threshold)
2. Writes tests first — never implementation first
3. Confirms tests fail (vacuous test guard)
4. Implements the minimum code to pass
5. Refactors with tests as a safety net
6. Loops until everything is green

Stack support

React/TypeScript (Vitest + RTL), Java/Spring Boot (JUnit 5 + Mockito),
Node.js/Express (Jest + Supertest), Playwright (multi-browser E2E),
Python/pytest, Go — auto-detected from project config files.

Claude Code extras

• Subagent phase isolation via Task() (test writer never sees implementation)
• PostToolUse hook for auto-running tests after every file save
• Config persisted to CLAUDE.md across sessions
• Monorepo support (Turborepo, Nx, Maven multi-module)

Gap this fills

There is an existing webapp-testing skill for Playwright, but no skill
that enforces TDD as a development loop. This works alongside webapp-testing.

Files

• SKILL.md — main skill with full 7-phase loop
• references/test-patterns.md — per-framework boilerplate
• references/ai-evals.md — LLM-as-judge harness for AI output testing
• references/claude-code-agents.md — subagent files for Claude Code

License

Apache 2.0