What problem does it solve? Reviewing pull requests thoroughly is hard: reviewers skim diffs, trust unverified claims like "tested locally", miss prior review threads, and rubber-stamp their own PRs. This Skill provides a structured reviewer-side workflow that catches what casual re-reads miss. ## Core Features & Use Cases - Self vs external review detection: Compares the PR author with your GitHub identity via gh to choose self-review or external-review mode, applying doubled rigor to your own code. - Full commit-range walkthrough: Reviews every commit oldest-first, flags mixed concerns, net-zero changes, and misleading commit messages, and recommends splitting oversized PRs instead of partial reviews. - Trust-but-verify pass: Treats every claim in PR descriptions and prior comments as unproven, constructing minimal falsifying inputs and checking backward compatibility for callers on old behavior. - Use Case: A teammate opens a 12-commit PR claiming a race condition fix. Use this Skill to walk each commit, build the adversarial case against the fix, check prior review threads for unaddressed feedback, and post one atomic inline review via the GitHub API. ## Quick Start Ask the AI to review pull request number 42 in the current repository using the giving-code-review workflow and show the findings before posting.
