Multi-Dimensional Code Review

What It Addresses

Code reviews can drift into style nitpicks or generic feedback. This skill constrains the review to five dimensions: code quality, security, performance, maintainability, and architecture & design, making it easier to catch security risks, hidden performance costs, and maintainability issues before merge. It is useful for reviewing a branch, a directory such as src/, or a single file.

How It Works

The skill is invoked through /code-review commands:
- /code-review: review all changes on the current branch
- /code-review src/: review a directory
- /code-review src/main.ts: review a file

The review checks naming, duplication, single responsibility, exception handling, dead code, and magic values. It also looks for SQL/NoSQL injection, XSS, SSRF, hardcoded secrets, dependency vulnerabilities, N+1 queries, resource leaks, event-loop blocking, missing tests, weak logging, and unclear module boundaries. Findings should include the language, a suggested code example, and severity: 🔴 high must be fixed, 🟡 medium should be fixed, and 🟢 low is optional.

Boundaries

It works from the provided source and visible context, so it does not replace runtime tests, dynamic analysis, or production incident diagnosis. For large repositories, narrow the scope by directory or file. Conclusions about security and performance still need validation against project constraints, deployment environment, and actual behavior.