What problem does it solve? Agent-generated E2E tests often pass today but break on the first refactor or fail to catch real regressions. This Skill drives an approved plan's browser-level phases one risk at a time, ensuring each Playwright test actually protects a named risk and fails when that risk materializes. ## Core Features & Use Cases - Risk-driven test generation: Picks browser-level risks from a test plan, maps the user flow via the accessibility tree, and generates one reviewed test per risk using a seed exemplar and E2E quality rules. - Five anti-pattern review: Checks every generated test for hallucinated assertions, brittle selectors, shared state, wait-for-time, and missing cleanup, then re-prompts by name. - Deliberate-break verification: Confirms each test goes red when the protected behavior is broken, proving the assertion is not decorative. - Use Case: After building a flashcard save feature, run the Skill on the plan's E2E phase to produce a Playwright test proving card data survives a full page reload, verified by temporarily breaking the save logic. ## Quick Start Run /10x-e2e with your change-id, for example "/10x-e2e save-session phase 6", to drive that plan's browser-level E2E phases against the running app.