This sets up real TDD loops using Claude Code's Stop hook, which fires after each response and can feed test failures back to Claude automatically when it exits with code 2. You drop a bash script in your project that runs tests, lint, and typechecking, then returns failures as stderr that Claude sees and fixes in the next iteration. No fake plugins or commands, just hook infrastructure that's already in Claude Code. Works well for feature development where you want tests to drive the work, less useful for exploratory coding or when your test suite is slow. The iteration counter prevents infinite loops, capping at 25 tries by default.
npx skills add https://github.com/alinaqi/claude-bootstrap --skill iterative-development