What problem does it solve? Verifying that a web app actually works after shipping code usually means manual clicking, screenshotting, and note-taking. This Skill automates report-only QA: it drives a real browser through your app, documents bugs with evidence, and computes a health score, while never touching your code. ## Core Features & Use Cases - Diff-aware testing: On a feature branch with no URL given, it analyzes the git diff, maps changed files to affected pages and routes, and verifies the branch's changes end-to-end in a browser. - Structured bug reports: Every issue gets screenshots, repro steps, severity, and category, written to .gstack/qa-reports/ using a report template, plus a weighted health score across console, links, visual, functional, UX, performance, and content. - Multiple modes: Full systematic exploration, 30-second --quick smoke test, and --regression mode that diffs results against a saved baseline.json to show fixed issues, new issues, and score delta. - Use Case: You just pushed a branch changing the checkout flow. Run the skill with no arguments; it detects your local dev server, tests the affected pages, and hands you a report with screenshots proving what works and what broke. ## Quick Start Ask the AI to run /qa-only on your current feature branch or against a URL like http://localhost:3000 and produce a bug report without changing any code.
