This is a structured debugging protocol for end-to-end test failures. It forces you to classify every failure into one of three buckets (flaky infrastructure, outdated test, or actual bug) before you touch any code, which prevents the common mistake of patching tests when the app is wrong or vice versa. The workflow is opinionated about boundaries: it won't let you change source code without a spec to back it up, and it requires unit tests before marking bug fixes complete. The taxonomy is sharp enough that you can actually apply it during a test run without overthinking. If you work on a codebase where e2e tests fail often and nobody's sure whether to trust them, this gives you a repeatable way to triage and fix without making things worse.
npx skills add https://github.com/0xbigboss/claude-code --skill e2e