What problem does it solve? Manually clicking through a web app to find bugs is slow and inconsistent, and developers often ship branch changes without verifying they actually work in the browser. This Skill acts as a QA engineer that systematically tests a web application like a real user and produces an evidence-backed report — without ever touching the source code. ## Core Features & Use Cases - Report-only testing: Explores pages, clicks interactive elements, fills forms, checks console errors, and documents issues with screenshot evidence and repro steps — but never fixes anything. - Multiple modes: Diff-aware mode automatically scopes testing to files changed on a feature branch; Full mode systematically explores every reachable page; Quick mode runs a 30-second smoke test; Regression mode diffs results against a saved baseline. - Structured output: Generates a report with a weighted health score (console, links, visual, functional, UX, performance, content, accessibility), severity-ranked issues, a Top 3 Things to Fix list, and a baseline.json for future regression comparisons. - Use Case: You just pushed a feature branch and want to verify it works before merging. Run the Skill with no arguments — it detects the branch diff, finds your local dev server, tests the affected pages in a real browser, and hands you a report with screenshots of anything broken. ## Quick Start Ask the AI to run a report-only QA pass on your app at http://localhost:3000 and save the results to .gstack/qa-reports/.
