from claude-code-setup25
A failing-test-first workflow for agents: write a minimal failing test, implement the smallest change to pass, and refactor while keeping tests green—used for f
This skill codifies a TDD workflow for coding agents: prioritize executable evidence over guesswork by driving changes with failing tests. It prescribes how to choose the smallest seam, pick fast verification commands, and run red->green->refactor cycles while keeping feedback loops fast and reliable. The guidance covers new features, bug reproduction, characterization tests for legacy code, and refactors.
Use when an agent needs high-confidence changes backed by tests: adding features, fixing bugs, reproducing regressions, or refactoring risky areas. Prefer this skill when you want proof (tests) rather than solely model reasoning—examples include prompts like "use TDD", "write the test first", or "add a regression test".
Designed for code-oriented agents and developer toolchains (Claude Code, Copilot-style agents, CLI-based codex workflows) that can run tests and edit repositories programmatically.
This skill has not been reviewed by our automated audit pipeline yet.
