What problem does it solve? Reviewing a large pull request by reading raw diffs is slow and makes it hard to know which UI states and routes actually need browser testing. This Skill turns the current PR or branch diff into a structured visual recap that maps what changed and where reviewers should focus. ## Core Features & Use Cases - Diff Collection and Prompt Generation: A bundled helper script resolves the PR (via gh) or an explicit base/head range, collects the diff and stats, and builds a recap prompt using the Agent-Native CLI. - Local-First Recap Authoring: The agent authors a plan.mdx recap covering changed surfaces, affected routes/components, before/after UI states, API or schema changes, risk notes, and a focused browser QA checklist. - Local Serving for Review: The recap is verified and served locally through Agent-Native local-files mode, producing a shareable local Plan URL without GitHub Actions or API keys. - Use Case: Before merging a frontend PR, invoke the skill to get a review map of changed components and a VAR/browser QA checklist, so reviewers test the right screens instead of scanning every changed line. ## Quick Start Ask the agent to run $local-visual-recap to generate a local visual recap for the current PR and serve it for browser QA review.
