What problem does it solve? Manual QA of a running web app is slow and inconsistent: edge cases, empty/error/loading states, and console errors get missed, and found bugs often lack reproducible evidence or verified fixes. This Skill drives a real browser through the app like a user would, documents every issue with before/after evidence, and fixes bugs with atomic commits and regression tests. ## Core Features & Use Cases - Systematic browser exploration: Uses the claude-in-chrome MCP tools to click every interactive element, fill and submit forms with invalid and edge-case inputs, walk navigation, and check the console after every interaction. - Multiple QA modes: Diff-aware testing of changed pages on a feature branch, full-site audits, quick smoke passes, regression comparison against a baseline, and a report-only mode with no fixes. - Evidence-backed fix loop: Each bug gets before/after screenshots or a recorded repro, a health score across eight categories, then a minimal fix in its own fix(qa): commit with a regression test and re-verification. - Use Case: After finishing a feature branch, ask for a QA pass; the Skill maps the diff to affected pages, tests them in your real Chrome, fixes three bugs with atomic commits, and reports a health score moving from 72 to 91. ## Quick Start Ask the agent to QA the running app on this feature branch and fix any bugs it finds with evidence.
