performance

What This Skill Does

The Performance skill empowers OpenClaw to act as a senior-level Performance Engineer. It provides a structured methodology for identifying, measuring, and resolving bottlenecks across the entire technology stack—from the browser's critical rendering path to database query execution plans. By leveraging established industry standards like the RAIL model and Core Web Vitals, this skill ensures that web applications remain responsive, scalable, and efficient. It guides the user through a disciplined workflow: measuring, analyzing, implementing one optimization at a time, and re-validating.

Installation

To integrate this performance suite into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/anton-abyzov/sw-performance

Use Cases

• Frontend Optimization: Identifying large bundle sizes using webpack-bundle-analyzer, implementing code-splitting, and optimizing image delivery to improve LCP and CLS scores.
• Database Tuning: Detecting N+1 query patterns in ORM layers, adding necessary indexes to foreign keys, and using EXPLAIN to refactor slow-running SQL queries.
• Distributed Systems & Backend: Implementing Redis-based caching layers for hot data, configuring connection pooling, and optimizing asynchronous processing to handle high-concurrency workloads.
• Load Testing: Baseling application performance under stress to determine service breaking points and potential failure modes.

Example Prompts

1. "My dashboard is loading in over 5 seconds. Can you help me profile the LCP and identify what's blocking the main thread?"
2. "I'm seeing a significant spike in database latency on our user report endpoint. Can you analyze my current query and help me resolve any potential N+1 issues?"
3. "Our bundle size is currently 2MB for the main entry point. Please guide me through a strategy to implement route-based code splitting using React.lazy."

Tips & Limitations

• Measure First: Never attempt an optimization without concrete telemetry. Always start by verifying the bottleneck.
• One Area at a Time: To avoid regression, focus on a single performance bottleneck per session. Don't mix frontend, database, and caching fixes in one PR.
• Automation Friendly: This skill is highly effective when paired with tools like k6, Lighthouse, and New Relic. Always provide the raw logs or report data to the agent for the most accurate diagnosis.