What problem does it solve? Code that matches its documentation can still fail in production: the spec is not the consumer, test layers do not enforce what browsers enforce, concurrency bugs hide from sequential calls, and tests that cannot fail produce false confidence. This Skill closes the gap between "correct on paper" and "actually works" by requiring real execution and state re-verification before declaring anything done. ## Core Features & Use Cases - Reality-based verification: Run flows exactly as the real consumer runs them (e.g., claude.ai, Telegram, Meta Graph API) rather than as the spec describes, and read the other side's logs. - Write verification via state re-reads: Treat a write call's return value as receipt of the request, not proof of the result; confirm every write, send, or delete by reading the state back. - Falsifiable tests: Require every test to have a failure scenario, and run bug-fix tests against the old code to watch them fail before trusting them. - Use Case: After fixing an endpoint bug, run the flow with curl against the live service, re-read the stored state to confirm the change, and execute the new regression test on the pre-fix code to prove it catches the bug. ## Quick Start Verify that the fix you just made actually works by running the flow end-to-end against the live service and re-reading the resulting state.
