---

name: code-review description: "Use when reviewing code for production readiness. Identifies critical issues causing production failures: data loss, security breaches, race conditions, performance degradation. Focuses on measurable impact, not style preferences. Triggers: reviewing code changes, verifying code safety/correctness, pre-commit reviews, after implementing significant changes."

Code Review Practices

Core Principles

• Measurable Impact: Focus on issues causing actual failures: data loss, security breaches, race conditions, performance degradation. Ignore theoretical problems without real impact.
• Production Perspective: Consider how code behaves under production load and error conditions.
• Actionable Feedback: Provide specific locations and clear reasoning for issues found.
• Balanced Scrutiny: Find critical issues without being pedantic about minor concerns.

Project Context

Check project documentation for: quality standards/conventions, error handling patterns, performance requirements, architecture decisions, domain concepts/terminology

Critical Issue Categories

Production Failures (highest priority)

Data Loss Risks:

• Missing error handling that drops messages or data
• Incorrect acknowledgment before successful write
• Race conditions in concurrent writes
• Data corruption scenarios

Security Vulnerabilities:

• Credentials in code or logs
• Unvalidated external input (SQL injection, command injection, path traversal)
• Missing authentication or authorization checks
• Sensitive data exposure

Note: Only suggest high-performance validation checks. Avoid expensive validation in hot paths without justification.

Performance Killers:

• Unbounded memory growth
• Missing backpressure handling
• Blocking operations in hot paths
• O(n²) algorithms where O(n) is straightforward
• Unnecessary allocations in tight loops

Concurrency Bugs:

• Shared mutable state without synchronization
• Thread or task leaks
• Deadlock conditions
• Race conditions

Degraded Operation (medium priority)

Correctness Issues:

• Logic errors affecting business requirements
• Incomplete error propagation
• Missing error recovery mechanisms
• State management issues

Resource Management:

• Resource leaks (connections, file handles, memory)
• Missing cleanup on error paths
• Background tasks that cannot be terminated

Maintainability Issues:

• Unnecessary complexity that obscures intent
• Code duplication that should be unified
• New functions duplicating existing patterns without justification
• Mixing abstraction levels inappropriately

Error Boundaries:

• Implementation details leaking through error types across abstraction boundaries
• Database errors exposed as API errors (should be transformed to domain errors)
• File system errors bubbling up unchanged (should be contextualized)

Domain Alignment:

• Using generic technical terms instead of domain concepts
• Code structure not reflecting domain boundaries
• Missing domain terminology in critical areas

Comments:

• Incorrect or outdated comments
• Obvious comments that restate what code does
• Missing comments for non-obvious decisions

Low Priority or Non-Issues

Avoid flagging:

• Style preferences without functional impact
• Theoretical edge cases with no realistic risk
• Minor optimizations with negligible benefit
• Alternative implementations that are equivalent
• Cosmetic concerns

Review Checks

Verify Error Handling: Check errors handled appropriately at each level. Errors should be transformed when crossing abstraction boundaries, not just propagated unchanged.

Example issue:

Copy# Problem: Database error leaks to API layer
def get_user(user_id):
    return db.query("SELECT * FROM users WHERE id = ?", user_id)
    # If query fails, DatabaseError bubbles up to API

# Better: Transform to domain error
def get_user(user_id):
    try:
        return db.query("SELECT * FROM users WHERE id = ?", user_id)
    except DatabaseError as e:
        raise UserNotFoundError(f"User {user_id} not found") from e

Check Concurrency Safety: Verify shared mutable state is properly synchronized. Look for race conditions in concurrent operations.

Example issue:

Copy# Problem: Unsynchronized shared state
class Worker:
    count = 0  # Shared mutable state

    def process(self):
        self.count += 1  # Race condition

Validate Resource Management: All resources should be properly closed or released. Cleanup must happen even on error paths. Background tasks should be terminable.

Check Domain Alignment: Code should use appropriate domain terminology. Does it reflect system's purpose and concepts? Are abstractions at right boundaries?

Review Complexity: Is complexity justified by requirements? Can code be simplified without losing functionality? Does control flow jump around unnecessarily? Are functions extracted appropriately (not too shallow)?

Verify Architecture Alignment: If architecture documentation or decisions exist, verify code aligns with them. Check for consistency with established patterns.

Consider Boy Scout Opportunities: Can nearby code be improved opportunistically alongside main change? Or does it require separate effort that should be deferred?

Guiding Principles

Prefer real issues over exhaustive coverage:

• Miss minor issues rather than flag non-issues
• Focus on actual production failures
• Consider cost-benefit of changes

Avoid absolutism:

• Not every error needs handling (sometimes failing fast is correct)
• Not every optimization is worth the complexity
• Not every duplication needs elimination

Context matters:

• Hot paths have different standards than setup code
• Prototype code has different standards than production code
• Core infrastructure has different standards than application code