bakeoff/live/v4/code-review/code-review/SKILL.md
Provides comprehensive code review assistance for developers analyzing pull requests and code changes. Use when performing code reviews, examining diffs, or evaluating code quality before merge. Covers bugs, security vulnerabilities, performance concerns, and maintainability with structured feedback and severity levels.
npx skillsauth add bjg4/bgskillz code-reviewInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
This skill helps developers review code effectively. Code review is a collaborative process where experienced developers examine changes to catch issues before they reach production.
Use cases:
Unstructured reviews miss critical issues because reviewers jump around without a system. Structured findings with severity levels help authors prioritize fixes. Explaining why something is a problem teaches the team and reduces repeat mistakes.
When reviewing code, work through these areas in order because security and correctness issues outweigh style nits:
Look for off-by-one errors, null handling, race conditions, and missing edge cases. Ask whether the change handles failure paths.
Examine input validation, authentication, authorization, secrets handling, and injection risks (SQL, XSS, command injection). Security issues should be flagged as high severity because they can cause real harm.
Consider naming clarity, function length, duplication, test coverage, and whether the change fits existing patterns. Maintainability affects long-term velocity.
Note obvious inefficiencies — N+1 queries, unnecessary allocations, blocking calls in hot paths — when relevant to the change.
Produce structured findings grouped by severity:
Each finding should include: location (file/line if available), description of the issue, and a suggested fix or direction.
Be constructive. Explain reasoning so the author learns. Avoid vague comments like "this is wrong" without saying why.
If the diff or code context is missing, ask the user to provide it. If the request is outside review scope (e.g., writing new features from scratch), explain that and offer to switch tasks.
Adapt review depth to the change size — a one-line typo fix needs a lighter pass than a new auth module. Match the user's stated priorities (security-heavy vs speed-heavy) when they indicate them.
Good finding: "Line 42: SQL query uses string interpolation with user input — SQL injection risk. Use parameterized queries."
Weak finding: "Security could be better."
development
Create S-tier portable skills and agents with comprehensive quality guidance. Use when creating or improving skills, designing multi-agent orchestration workflows, auditing agent quality, running evals, or learning agent-building best practices. Covers the full lifecycle from use case definition through automated evaluation, review, and iteration. Do NOT use for general coding tasks, writing documentation unrelated to agents/skills, or building applications.
development
Create S-tier portable skills and agents with comprehensive quality guidance. Use when creating or improving skills, designing multi-agent orchestration workflows, auditing agent quality, running evals, or learning agent-building best practices. Covers the full lifecycle from use case definition through automated evaluation, review, and iteration. Do NOT use for general coding tasks, writing documentation unrelated to agents/skills, or building applications.
development
Create S-tier portable skills and agents with comprehensive quality guidance. Use when creating or improving skills, designing multi-agent orchestration workflows, auditing agent quality, running evals, or learning agent-building best practices. Covers the full lifecycle from use case definition through automated evaluation, review, and iteration. Do NOT use for general coding tasks, writing documentation unrelated to agents/skills, or building applications.
development
Stress-test a plan or design through Socratic questioning until shared understanding. Use when the user says "grill me", wants to stress-test a plan, or review a design. Do NOT use for implementation or code generation.