← ClaudeAtlas

devtools-ux-auditlisted

Browser-evidence workflow for auditing a normal user's experience across desktop-first, responsive, and mobile web interfaces. Use when a UX review needs viewport-specific navigation, interaction, snapshot, console, network, or screenshot evidence; prefer Chrome DevTools and allow an equivalent browser tool only when it preserves the evidence contract.
bohewu/agents_pipeline · ★ 1 · Web & Frontend · score 77
Install: claude install-skill bohewu/agents_pipeline
# DevTools UX Audit Collect browser evidence for one bounded user journey. This is the evidence layer; pair it with `$run-ux` when the user wants a formal multi-perspective scorecard and synthesized report. Do not use it as the primary workflow for native interfaces that cannot be exercised through a browser. ## Capability Preflight Before auditing: 1. Identify the available browser capability. Prefer Chrome DevTools tools. An equivalent browser or Playwright workflow is acceptable only if it can preserve viewport, interaction, snapshot, console/network, and screenshot evidence relevant to the task. 2. Confirm the target URL is reachable. 3. Decide who owns the local server and browser session. 4. Select a product profile and viewport preset. If no suitable browser capability is available, do not infer rendered behavior from source alone. Report the evidence limitation and either stop or perform a clearly labeled source-only review when that still helps. For Chrome-specific call order and recovery, read `references/chrome-devtools.md`. For Windows lifecycle details, read `references/windows-notes.md`. ## Ownership and Cleanup - A paired server workflow owns startup, readiness, and teardown of a local preview. - This task owns only the processes, pages, profiles, or browser sessions it starts. Never kill an unrelated shared browser or server. - Reuse an already-authorized page or session when practical. - If this task starts a server, completion requires the URL to st