Skill v1.0.0

version: "1.0.0" name: code-review-and-quality description: Conduct staff-engineer-level code reviews across correctness, security, performance, readability, and architecture difficulty: staff domains: [general]

Overview

Code review is the most important quality gate in software development. A good code review catches bugs, prevents security vulnerabilities, improves design, and transfers knowledge — before code reaches production. This skill runs a structured review across five dimensions.

When to Use

• Before any PR is merged
• When reviewing AI-generated code
• As part of the /review workflow

Process

Dimension 1: Correctness

• Does the code do what the spec says?
• Are all edge cases handled? (empty inputs, null, zero, max values, concurrent access)
• Are error conditions handled? (network failure, invalid input, resource exhaustion)
• Are there off-by-one errors?
• Are there race conditions in concurrent code?
• Does the code handle partial failures?

Dimension 2: Security

• Is all user input validated before use?
• Are all database queries parameterized?
• Is no sensitive data logged?
• Are all auth/authz checks correct?
• Are there injection vulnerabilities? (SQL, shell, LDAP, XML)
• Are there new attack surfaces (new endpoints, new data types)?
• See security-and-hardening for full checklist.

Dimension 3: Performance

• Are there N+1 query patterns?
• Are there missing database indexes?
• Is expensive computation done in a hot path?
• Are large data sets loaded into memory when streaming would work?
• Are there blocking synchronous calls that should be async?
• Does the change affect the critical path? Profile it.

Dimension 4: Readability and Maintainability

• Does the code express its intent clearly?
• Are names accurate? (Function names that lie are worse than no names)
• Is complexity necessary, or is there a simpler solution?
• Are there long functions that should be extracted?
• Is there duplicated logic that should be shared?
• Are there magic numbers or strings that should be constants?
• Will a new engineer understand this in 6 months?

Dimension 5: Architecture

• Does this change fit the existing design?
• Does this create new coupling that will be hard to remove?
• Does this introduce a new pattern when an existing one could be used?
• Does this create a new dependency that adds risk?
• Does this make the system harder or easier to test?

Review process

1. Read the PR description. If it's missing, ask for it.
2. Understand the spec/ticket the PR is implementing.
3. Review the tests first — they tell you what the author thinks the code should do.
4. Review the implementation against each dimension above.
5. Run the code locally for non-trivial PRs.
6. Leave specific, actionable comments. Not "this is wrong" — "this will fail when X because Y; consider Z."

Anti-Rationalizations

"The tests pass — it's good to go" Tests passing proves the code passes the tests. It does not prove the tests are correct, complete, or that the code is secure.

"It's a small change — it doesn't need a full review" Small changes have caused large incidents. The review checklist is short enough to run on any change.

Verification Requirements

• [ ] All five dimensions reviewed
• [ ] Comments are specific and actionable (not vague)
• [ ] Security dimension completed for any change touching auth, data, or external APIs
• [ ] Tests reviewed before implementation