performance-engineer

What This Skill Does

The performance-engineer skill serves as an elite technical companion for developers and SREs tasked with identifying, diagnosing, and resolving application latency and throughput bottlenecks. It provides a structured methodology for full-stack optimization, moving from initial observability and baselining to deep-code performance tuning. The skill leverages industry-standard tools such as OpenTelemetry for distributed tracing, enabling end-to-end visibility across microservices. It facilitates the implementation of multi-tier caching strategies, database query optimization, and Core Web Vitals remediation to improve both backend responsiveness and frontend perceptual performance. Beyond code-level fixes, it integrates load-testing frameworks like k6 and JMeter to validate architectural changes against simulated production traffic, ensuring your infrastructure is built to scale.

Installation

To integrate this skill into your environment, use the OpenClaw CLI within your project root: clawhub install openclaw/skills/skills/anton-abyzov/sw-performance-engineer

Use Cases

• Production Latency Spikes: Use this skill when your dashboard alerts show increased P99 response times, allowing the AI to trace the request lifecycle and identify blocking I/O or inefficient algorithms.
• Frontend Optimization: Utilize the skill to audit and improve Core Web Vitals scores for better SEO and user retention.
• Scalability Audits: Before a major launch or high-traffic event (like Black Friday), use the skill to set up stress tests and identify the breaking point of your services.
• Caching Strategy: Implement sophisticated multi-tier caching (Redis, CDN, Browser cache) to reduce expensive database hits.

Example Prompts

1. "My API latency has jumped from 200ms to 800ms under moderate load. Can we set up OpenTelemetry to trace the culprit?"
2. "I need to implement a Redis caching layer for our user profile microservice. What is the optimal eviction policy and structure for this?"
3. "Run a k6 load test script against our staging environment simulating 500 concurrent users to identify where our current bottlenecks are."

Tips & Limitations

• Incremental Delivery: For complex projects exceeding 1000 lines of implementation code, the AI will default to an incremental delivery mode. Always follow the prompts to proceed component-by-component (e.g., Monitoring → Caching → Database → Load Testing) to ensure high-quality code generation.
• Context is Key: Always provide the AI with your stack (e.g., Node.js, Go, Postgres, Redis) so the generated configurations are compatible with your existing infrastructure.
• Environment Safety: Never point aggressive load testing scripts at your production database without proper safety precautions and read-only replicas.