What problem does it solve? Running browser-based UI verification inside the main agent session floods its context with screenshots and DOM state, leaving little room for actual fixing work. This Skill enforces a delegation pattern that keeps visual QA out of the main context while still producing repeatable, reviewable test plans. ## Core Features & Use Cases - Repeatable Test Plans: Requires writing a concrete test plan (user goals, steps, expected results) before any browser walk begins, with human review for high-stakes changes. - Subagent-Driven Walks: A Sonnet subagent owns the entire navigate → screenshot → judge loop, keeping screenshots and DOM in its own context. - Concise Reporting: The main agent receives only pass/fail per goal, findings, and screenshot paths, then fixes issues and re-delegates for re-verification. - Use Case: After shipping a UI change, hand the subagent the staging URL and test plan; it reports which goals pass or fail, and the main agent patches the code and re-runs the walk. ## Quick Start Ask the agent to verify the updated checkout flow in the browser by delegating a QA walk with a written test plan to a subagent.
