do-design-auditlisted

Audit an existing web interface against premium design criteria and produce the structured report defined below — every rubric row scored with a specific finding, plus the three most impactful fixes. This is the review-time companion to `/frontend-design`: that skill builds to the standard; this one audits against it. Be opinionated. Call out what fails clearly. "Acceptable" is not a compliment. ## Repo Context Probe If `.claude/skill-context/do-design-audit.md` exists, read it and honor its declarations; otherwise use the generic defaults described below. The context file is where a repo declares how its session harness opts a session into real-Chrome mode (the exact CLI flag / config it uses). When the file is absent, the generic baseline applies: ensure your browser-automation session has whatever "use the real, logged-in browser" mode your harness provides before running the audit. ## Surface This skill runs against the user's real, logged-in Chrome via the BYOB MCP server (`mcp__byob__browser_*`). Public marketing pages and authenticated dashboards are audited the same way — BYOB just shows you the page the user would see. There is no anonymous-headless fallback. The calling session **must** be in real-Chrome mode (the harness may auto-infer this from the request, or expose an explicit flag — see the repo context file for the project-specific mechanism). Two concurrent real-Chrome sessions race on the active tab and corrupt each other's DOM. ## Variables - `url`