What problem does it solve? Code that was written is not code that runs. This Skill closes the gap between "implementation complete" and "actually works" by starting the dev server, running the test suite, exercising each scenario in a real browser, and recording per-scenario evidence in a verification.md file — without ever fixing what it finds. ## Core Features & Use Cases - Runtime command resolution: Discovers how to start the server and run tests from plan.md, spec/architecture.md, or the project manifest (package.json, pyproject.toml, Makefile, Cargo.toml, and more), never guessing commands. - Test suite execution with honest reporting: Records the exact command, real failure output, passed/failed/skipped totals, and duration; names skipped tests individually and maps each scenario to the test that actually covers it. - Browser scenario verification: Drives scenarios from scenarios.feature via Playwright, Cypress, a browser MCP, or a scripted manual pass, capturing screenshots and console errors. - Strict no-fix invariants: Never edits code, never writes inside spec/, and never installs browser automation stacks — failures are recorded and handed back to the implementation phase. - Use Case: After a cycle's tasks are checked off, ask it to verify the cycle; it starts the server, runs the tests, exercises each scenario, and writes verification.md as evidence for the validation phase. ## Quick Start Verify the latest cycle by running the dev server, executing the test suite, and exercising each scenario in the browser, then write the results to verification.md.
