← ClaudeAtlas

browser-qalisted

Browser automation for visual layout review or QA verification. Use only when the human explicitly names a page or check to run. Never run unsolicited or assume a broad test suite is wanted.
davidleberknight/footbag-platform · ★ 0 · Web & Frontend · score 65
Install: claude install-skill davidleberknight/footbag-platform
# Browser QA / Visual Layout Use this skill only when the human explicitly asks to look at a page, check a layout, or verify rendered output. Do not run it speculatively, do not expand scope beyond what was named, and do not take screenshots unless specifically useful for the stated goal. This skill is for ad-hoc human-driven browser review. The standing Playwright test suite (lightweight at `tests/e2e/`, heavyweight invoked separately) is described in `docs/TESTING.md` §6 and is a separate concern from this skill. ## Two modes **Visual layout review**: human wants to see how a page looks and give feedback. - Navigate to the named page, take one screenshot, report what is rendered. - Do not run assertions or check the full target rendering invariants (the view-layer standard and the page's service-JSDoc contract) unless asked. **QA verification**: human wants to confirm specific behavior is correct. - State exactly what will be checked before running anything. - Only check the URLs and interactions the human named. - Do not expand to adjacent pages or broad QA passes. ## Token discipline - Use `mcp__playwright__browser_snapshot` (accessibility tree) for content/structure checks: cheaper than screenshots. - Use `mcp__playwright__browser_take_screenshot` only when visual appearance is the explicit goal. - Do not take multiple screenshots per page unless the human asks for it. - Close the browser when done (`mcp__playwright__browser_close`). ## Screenshot file hygiene *