Skip to content

charlesjones-dev/ai-performance

v2.0.0MIT

Audits a codebase, or one directory, for performance bottlenecks and saves a timestamped report with an impact score for each finding

GitHub

AI-Performance Plugin

Performance audits for Claude Code, Codex and Cursor. /performance-audit reads your codebase and saves a timestamped report with an impact score for each bottleneck it finds.

What This Plugin Does

The plugin ships one skill. Run the audit during development to catch code-level performance problems before they reach production.

Limits

/performance-audit reads code. It doesn't run your app, so it can't measure execution time or Core Web Vitals in a browser; pair it with profiling and load tests. Improvement figures in the report are estimates.

Available Skills

/performance-audit

Audits the whole repository by default. To limit the audit to one part of the code, pass a directory path relative to the repository root, for example /performance-audit src/api. The report still goes to docs/performance/ at the repository root. Other text after the command is ignored.

The audit asks no questions while it runs, so you can use it for unattended or scheduled reviews.

Quick Start

Installation

Claude Code:

/plugin marketplace add charlesjones-dev/ai-agent-plugins
/plugin install ai-performance@ai-agent-plugins

Codex:

codex plugin marketplace add charlesjones-dev/ai-agent-plugins
codex plugin add ai-performance@ai-agent-plugins

Cursor: add this repository as a team marketplace, or copy this folder to ~/.cursor/plugins/local/ai-performance. See the Cursor install steps.

Usage

# Audit the whole repository
/performance-audit

# Audit one directory
/performance-audit src/api

# Review the generated report in docs/performance/

Features

/performance-audit

Performance Pattern Detection

  • N+1 Query Problems: Identifies loading related entities in loops
  • Synchronous/Blocking Operations: Detects blocking database calls, I/O operations, and event loop blocking (Node.js)
  • Memory Issues: Finds leaks from closures, uncleared listeners, large object allocations, and missing resource disposal
  • Concurrency Problems: Detects race conditions, deadlocks, connection pool exhaustion, and thread/worker saturation
  • Inefficient Queries: Detects suboptimal ORM usage and missing projections
  • Missing Caching: Locates frequently computed operations without caching
  • Database Optimization: Identifies missing indexes and expensive queries
  • Frontend Performance: Evaluates Core Web Vitals impact (LCP, INP, CLS), bundle size, and rendering performance
  • GraphQL Issues: Detects missing query depth limiting, DataLoader opportunities, and over-fetching

Detailed Reporting

Each finding includes:

  • Location: Exact file path and line number
  • Performance Impact: Numerical impact score (1.0-10.0) with defined severity thresholds
  • Pattern Detected: What performance issue was identified
  • Code Context: The problematic code snippet
  • Impact: Performance cost and scalability concerns
  • Recommendation: How to optimize it
  • Fix Priority: When it should be addressed

Code Optimization Examples

Reports include 2-4 before/after examples built from code in the findings, each with an expected improvement estimate.

How It Works

/performance-audit scans source files for the patterns above, only under the given directory if you pass one, and tailors findings to the detected stack. Findings are grouped as Critical, High, Medium or Low, and the report ends with a phased optimization roadmap.

Best Practices

When to Run Performance Audits

  • Before production deployments
  • After implementing new features with database access
  • When adding new API endpoints
  • After integrating third-party services
  • During performance reviews and optimization sprints

How to Use Audit Results

  1. Prioritize Critical & High findings - Address these immediately
  2. Review code context - Understand why each finding impacts performance
  3. Apply optimization examples - Use provided code fixes as templates
  4. Measure improvements - Benchmark before and after optimizations
  5. Re-audit after fixes - Verify improvements and track progress

Configuration

There's nothing to configure.

Reports are saved to docs/performance/ at the repository root as YYYY-MM-DD-HHMMSS-performance-audit.md (e.g., 2025-10-17-143022-performance-audit.md). This timestamp-based naming ensures multiple audits on the same day don't overwrite each other.

Plugin Details

  • Name: AI-Performance
  • Version: 2.0.0
  • Type: Skill
  • Features:
    • Skills: /performance-audit
  • License: MIT
  • Author: Charles Jones

Contributing

Bug reports and fixes are welcome. See CONTRIBUTING.md.

License

MIT License - See LICENSE file for details.