code-review

name	code-review
description	Post-implementation code review methodology with severity framework and project-specific rules. Use when reviewing code changes, auditing pull requests, analyzing code quality, or performing security review.

Code Review

Structured methodology for reviewing code changes — security, correctness, quality, and project convention adherence. Produces actionable findings with severity classification and a clear verdict.

---

When to Use This Skill

• Reviewing code after implementation (post-build audit)
• Analyzing pull requests or staged changes
• Security-focused code audit
• Checking adherence to project conventions
• Validating that generated code wasn't manually edited

---

Methodology

Phase 1: Scope & Context

1. Identify the change set — files modified, lines changed, modules affected
2. Read surrounding code — understand the context before judging the change
3. Check module boundaries — are changes confined to one module or cross-cutting?

Phase 2: Review Checklist

Apply each category systematically. Skip categories that don't apply to the change.

Security

Check	Look For
Injection	SQL, command, XSS vulnerabilities
Auth	Missing auth checks, token handling
Secrets	Hardcoded credentials, exposed keys
Validation	Missing input validation

Correctness

Check	Look For
Logic	Off-by-one, null checks, edge cases
Types	Type safety, proper generics usage
Error handling	Unhandled exceptions, error propagation
Tests	Missing tests for new behavior

Quality

Check	Look For
Patterns	Consistency with codebase conventions
Complexity	Overly complex solutions, unnecessary abstractions
Documentation	Missing docstrings, outdated comments
Dead code	Unreachable or unused code

Project Rules

Check	Look For
Generated code	Manual edits to *_generated/ directories
Types	Use of any (TS) or Any (Python)
Commands	Direct npm/poetry usage instead of make targets
Module boundaries	Cross-module imports violating isolation
Statelessness	Instance variables storing request-scoped data

Phase 3: Severity Classification

Classify each finding:

Severity	Criteria	Action
Critical	Security vulnerability, data loss risk, production crash	Must fix before merge
High	Type safety violation, missing error handling, broken contract	Should fix before merge
Medium	Missing tests, inconsistent patterns, minor quality issue	Fix recommended
Low	Style suggestions, minor improvements, nice-to-haves	Optional

Phase 4: Verdict

Verdict	When
✅ Approve	No Critical/High findings
⚠️ Approve with comments	Only Medium/Low findings
❌ Request changes	Any Critical or High findings

---

Output Format

## Code Review: [Scope]

### Summary
[1-2 sentence overview of changes reviewed]

### Findings

#### 🔴 Critical
| Location | Issue | Recommendation |
|----------|-------|----------------|

#### 🟠 High
| Location | Issue | Recommendation |
|----------|-------|----------------|

#### 🟡 Medium
| Location | Issue | Recommendation |
|----------|-------|----------------|

#### 🔵 Low / Suggestions
| Location | Issue | Recommendation |
|----------|-------|----------------|

### Positive Observations
- [Good patterns observed]

### Statistics
- Files reviewed: X
- Critical: X | High: X | Medium: X | Low: X

### Verdict: [✅ | ⚠️ | ❌]

---

Anti-Patterns

• ❌ Reviewing without reading surrounding code — leads to false positives
• ❌ Style-only feedback disguised as High severity — undermines trust
• ❌ Asserting "no tests exist" without searching — verify before claiming absence
• ✅ Every finding has a specific file:line reference and actionable recommendation