What problem does it solve? Reviewing a large code change without guidance is slow and error-prone — reviewers miss intent, skip high-risk spots, and never verify behavior. This Skill turns an AI agent into a review facilitator that walks a human through a change step by step, from orientation to a final approve/rework decision. ## Core Features & Use Cases - Guided Review Trail: Builds or reuses a Suggested Review Order of path:line stops grouped by design concern, so the reviewer reads code in comprehension order rather than diff order. - Risk Surfacing: Scans the diff for high-blast-radius spots (auth, schema, billing, security, public APIs) and presents them as clickable path:line references, plus targeted "dig into [area]" deep-dives on demand. - Experiential Testing Suggestions: Proposes concrete manual observations (commands to run, responses to expect) instead of duplicating CI checks. - Use Case: After finishing a feature branch, say "checkpoint" — the agent identifies the change from git state or sprint status, summarizes intent and surface-area stats, walks you through the design concerns, flags risky spots, suggests how to see it working, and helps you approve the PR via gh pr review or plan a rework. ## Quick Start Say "walk me through this change" or "checkpoint" after completing work on a branch to start a guided human review of the diff.
