What problem does it solve? Manually verifying that a pull request's UI changes actually work in a browser is slow and often skipped, letting visual regressions and broken states slip through review. This Skill automates that verification: it boots the app, drives the changed screens through a real browser, captures screenshots, and produces an honest pass/fail report without ever modifying source code. ## Core Features & Use Cases - Browser-based UI verification: Derives a scoped QA scenario from the PR diff, exercises the changed routes through a configured browser provider, and captures a screenshot at every checkpoint. - Review-first gating and PR integration: Ensures the PR has been code-reviewed first (invoking om-auto-review-pr when needed), claims the PR with a lock, and posts evidence inline as a PR comment with optional self-QA sign-off labels. - Local tracker-less mode: Runs against the current worktree without any tracker, writing report.json, report.md, and screenshots to a local artifacts directory. - Use Case: A developer opens a PR changing a checkout form. The skill checks out the PR head in an isolated worktree, boots the app via om-prepare-test-env, walks the form's default, error, and mobile states in a browser, and posts a screenshot-backed PASS/FAIL evidence report as a PR comment. ## Quick Start Ask the agent to run UI QA on pull request 123 with om-auto-qa-pr and post the screenshot evidence as a PR comment.
