Enhance Code Review Quality

Code Review Excellence is a skill for conducting systematic, constructive code reviews that turn review into knowledge sharing rather than gatekeeping. It structures the review into four phases: context gathering (reading the PR description, checking size, CI status, and business requirement), a high-level pass (architecture and design fit, file organization, testing strategy), a line-by-line pass (logic and correctness, security, performance, maintainability), and a summary/decision step (approve, comment, or request changes). It packages four review techniques - a checklist method covering security, performance, and testing; the question approach for surfacing issues as questions instead of statements; suggesting instead of commanding; and differentiating feedback by severity (blocking, important, nit, suggestion, learning, praise). It includes language-specific review patterns for Python (mutable default arguments, overly broad exception handling, mutable class attributes) and TypeScript/JavaScript (use of any, unhandled async errors, prop mutation), plus advanced patterns for architectural review (staged PRs, design-doc-first for large features), test-quality review (behavior testing versus implementation-detail testing), and a security review checklist covering authentication, input validation, and data protection. It also gives a context-plus-specific-issue-plus-helpful-solution pattern (a modified sandwich method) for delivering difficult feedback and a five-step approach for handling author disagreement, from seeking to understand the author's reasoning through escalating to a senior reviewer when needed. Best-practice guidance caps effective review sessions at 60 minutes and PR size at 200-400 lines, calls for reviewing within 24 hours, and names common pitfalls to avoid: perfectionism over minor style, scope creep, inconsistent standards, delayed reviews, ghosting after requesting changes, rubber-stamping approvals without actually reviewing, and bike-shedding over trivial details.

Use it when reviewing pull requests and code changes, establishing code review standards, mentoring developers through review feedback, or auditing for correctness, security, or performance. Do not use it when there are no code changes to review, the task is a design-only discussion without code, or the job is to implement fixes rather than review them.

Input is a pull request or code change plus its linked issue and CI status. Output is structured review feedback: a high-level summary of findings, issues grouped by severity (blocking, important, minor), suggestions and questions, and test and coverage notes, delivered through a PR review comment template with Summary, Strengths, Required Changes, Suggestions, Questions, and Verdict sections.

The skill is tool-agnostic in its guidance but references GitHub and GitLab as review platforms, and recommends pairing automated linters and formatters, naming Prettier and Black, with the manual review so reviewers focus on logic, security, and design rather than formatting.

Engineers and tech leads who review pull requests, run architecture reviews, or mentor other developers through code review feedback, and teams looking to establish consistent review standards.