What problem does it solve? Code written by an AI session is often reviewed by the same session, which rationalizes its own choices instead of scrutinizing them. This Skill runs a full pull-request review in a clean context — fetching the PR, running the project's validation suite, analyzing the diff, categorizing issues by severity, and posting an approve / request-changes / comment verdict to GitHub — so a human receives a triaged, validated PR to make the final call. ## Core Features & Use Cases - Fresh-eyes diff review: Dispatches a code-reviewer subagent (or reviews in a clean session) covering correctness, type safety, security, performance, tests, and standards compliance, reporting high-confidence issues only. - Validation gate: Runs the project's real test, type-check, lint, and build commands and treats a red suite as a finding in itself. - Severity triage and GitHub posting: Categorizes every issue as Critical / High / Medium / Low, writes a report to .claude/code-reviews/pr-{N}-review.md, and posts it via gh pr review. - Use Case: After piv-create-pr opens a PR, run this Skill on the PR number to get a documented, severity-ranked review posted to GitHub before a human merges. ## Quick Start Ask the agent to review pull request number 42 and post the verdict to GitHub.
