ISSUE 06
LIVEClaude CodeAI AgentsTesting
CONTEXT
New apps ship with big test coverage gaps and accumulated UX debt.Audits of either are manual, slow, and inconsistent.
BUILD
A Claude Code skill with two modes.In `code` mode it baselines the test infrastructure, maps every interaction, catalogs existing coverage in parallel, prioritizes uncovered paths by risk, then dispatches 3-5 parallel agents to write tests in rounds: pure functions first, stateful code second.In `ui` mode it tours the running app in a real browser via Playwright MCP, drives every control, captures screenshots, and sorts findings into bugs, friction, and dead features.Stack-agnostic; works on Next.js, React Native, Django, Rails, and Go.
RESULT
Took a real Expo app from 224 to 474 tests in a single session.A separate `ui` mode run on a React+Express app surfaced 15 bugs and 3 dead pages in one sitting.Distributed as an installable Claude Code plugin.
LINKS
SCREENSHOTS
