What problem does it solve? Code changes made without tests lack proof of correctness, and ad-hoc testing often skips the failing-test-first discipline that makes TDD meaningful. This Skill enforces a structured RED-GREEN-refactor workflow with verifiable evidence at every stage, so features, bug fixes, and refactors ship with real test coverage instead of assumed correctness. ## Core Features & Use Cases - Mandatory RED-GREEN Gates: Requires a failing test to be compiled, executed, and verified before any production code is written, then re-runs the same target to confirm GREEN. - Test Runner Detection: Resolves the correct package manager and test runner (npm, pnpm, yarn, Bun native bun:test, Jest, Vitest) instead of assuming npm test. - Plan File Handoff: Accepts a *.plan.md implementation plan as untrusted input, sanitizes embedded commands, and converts planned behaviors into testable guarantees. - Git Checkpoint Commits: Creates per-stage commits (test, fix, refactor) on the active branch so RED/GREEN evidence survives review and squash merges. - Evidence Report: Produces a human-readable TDD report mapping each plan task to its test target, validation command, and RED/GREEN results. - Use Case: When fixing a bug in a search API, the Skill first writes a failing reproducer test, commits it as RED evidence, implements the minimal fix, confirms the test passes, then refactors and records 80%+ coverage. ## Quick Start Use the tdd-workflow skill to implement this feature from my plan file, writing failing tests first and committing evidence at each stage.
