ux-reviewlisted

# UX review Checks **shipped UI implementation** against design bar — real-browser render, five objective gates (steps 4–8): a11y, tokens, responsive/render, states, regression vs baseline. Repo's _own_ tooling discovered, never assumed. Not code review (`/review`); not authoring/restyling (designer's scope). On request **orchestrates** remediation — one solver per blocker, re-render before merge (step 11). On invoke: brief user — render + check plan, fixes opt-in (per-blocker solvers, re-render before merge); gather inputs. Whatever the target's phrasing — a UI to check, or a question about ux-review itself — it's the surface to render; never substitute a hand-written explanation for actually rendering and gating below. ## Operating loop 1. **Ground the target and baseline.** Discover repo's _own_ tooling first (`CLAUDE.md`/`AGENTS.md`, README, CI config, dependency manifest): design-token source of truth, stories, a11y auditor, browser driver, visual-regression harness + invocation. By stack — tokens: design-token JSON/YAML, Style Dictionary, Tailwind/theme config, CSS custom properties, Figma MCP / Dev Mode server; stories: Storybook, Ladle, Histoire; a11y: axe-core, `pa11y`, Lighthouse; drivers: Playwright MCP tools, Cypress, devtools; visual regression: Playwright `toHaveScreenshot`, Chromatic, Percy, BackstopJS, reg-suit. Web-perf skill on host → page-load/Core-Web-Vitals go there, not here. Render target defaults to the invocation argument (`/ux-review <target>`) i