name: performance-review description: This skill should be used when the user asks to "check performance", "review for performance issues", "find slow code", "optimize", "check for memory leaks", "find N+1 queries", "check complexity", "profile code", "latency issues", or mentions improving code performance. version: 3.1.4

Performance Code Review Skill

Identify algorithmic inefficiencies, memory leaks, database query problems, and other performance bottlenecks through targeted performance-focused code review.

Agent Configuration

Uses performance-agent (Opus in thorough mode, Sonnet in gaps mode). See ${CLAUDE_PLUGIN_ROOT}/shared/review-workflow.md for authoritative model configuration.

Performance Categories Checked

Algorithmic Complexity (Major to Critical):

• O(n^2) nested loops that could be O(n)
• Repeated expensive operations inside loops
• Unnecessary full collection iterations
• Inefficient sorting or searching

Database Performance (Critical):

• N+1 query problems (query inside loop)
• Missing indexes on filtered/joined columns
• Unbounded SELECT queries
• Inefficient joins or subqueries

Memory Issues (Critical):

• Memory leaks from unreleased resources
• Growing caches without eviction
• Event listeners not removed
• Large object allocations in hot paths

Async/Await Patterns (Major):

• Sequential awaits for independent operations
• Blocking event loop with sync operations
• Missing Promise.all for parallel work
• Floating promises (no await or catch)

Language-Specific Issues:

Node.js/TypeScript:

• Blocking event loop with sync I/O
• String concatenation in loops
• Unnecessary array copies (spread in loops)

.NET/C#:

• Boxing/unboxing in hot loops
• LINQ allocations in frequently called code
• async void instead of async Task
• Missing ConfigureAwait in libraries

Auto-Validated Patterns

High-confidence patterns that skip validation. For full definitions, see ${CLAUDE_PLUGIN_ROOT}/shared/validation-rules.md.

Performance patterns: n_plus_one_query, nested_loop_includes, select_star_in_loop

Scope Prioritization

Hot Path Identification

Automatically prioritize:

• Files with "api", "handler", "controller", "route" in name (entry points)
• Files with "service", "repository", "dao", "query" in name (data access)
• Files with "util", "helper", "common" in name (frequently called)

Related Database Code

For N+1 and query performance detection, also read:

• ORM model definitions (models/*.ts, entities/*.cs)
• Repository implementations
• Migration files (to understand indexes)

False Positives

Apply all rules from ${CLAUDE_PLUGIN_ROOT}/shared/false-positives.md.

Performance-specific additions - do NOT flag:

• O(n^2) on small bounded data (max 10 items)
• Startup/initialization code (one-time cost)
• Code with caching/memoization elsewhere
• Theoretical improvements requiring benchmark proof

Complexity Analysis

Always include when relevant:

• Current: What is the Big O now?
• Optimized: What could it be?
• Data Scale: At what data size does this matter?

Example: "Current O(n*m) where n=orders, m=items. With 100 orders and 1000 items, this executes 100,000 comparisons. Optimized O(n+m) would be 1,100 operations."

Example Output

See examples/example-output.md for a sample showing:

• N+1 query with diff fix and complexity analysis
• Sequential awaits with Promise.all fix
• Unbounded query with pagination prompt

Related Components

See ${CLAUDE_PLUGIN_ROOT}/agents/performance-agent.md for the agent definition.