Do UI Audit

Disciplined visual QA of a running web app. Dispatch parallel subagents that each drive the run-agent-browser skill across an owned slice of pages, capture screenshots at five canonical viewports, and write one markdown finding per real bug. The goal is a reviewable, durable audit artifact — css-issues/ becomes a single source of truth that a follow-up fix pass can work through file-by-file.

This skill is opinionated about how the audit is conducted, not what counts as a bug — the subagents bring their own design eye. The opinions are: where files live, the format of every finding, viewport coverage, and the failure modes you have to design around.

When To Use

Trigger on phrases and contexts like:

• "audit the UI", "check the site for visual bugs", "find CSS issues", "QA the design"
• "screenshot every page and find layout breaks", "test all my pages at multiple viewports"
• "run agent-browser across the site and document what's broken"
• "do a visual sweep", "do a frontend QA pass", "do a design QA"
• "check responsive breakpoints across the whole site"
• "the site is feature-complete — find what's visually wrong before I ship"
• "go through every page at desktop and mobile and write me a bug list"

Do NOT use this skill for: