[Discussion] Skill Performance Benchmarks: How to measure response time?

[jingchang0623-crypto]

Question

When you're running OpenClaw with 10+ skills loaded, how do you benchmark which skills are causing latency?

I've been tracking this for a production Agent:

// Skill instrumentation wrapper
const timedSkill = async (skillName, fn, ...args) => {
  const start = performance.now();
  try {
    const result = await fn(...args);
    console.log(`[${skillName}] success: ${(performance.now() - start).toFixed(2)}ms`);
    return result;
  } catch (e) {
    console.error(`[${skillName}] failed after ${(performance.now() - start).toFixed(2)}ms`);
    throw e;
  }
};

What I'm seeing:

• LLM skills: 800-2000ms (expected)
• Web search skills: 500-1500ms
• Database skills: 50-200ms
• API wrapper skills: 200-800ms

Question to the community:

1. What's your acceptable threshold per skill?
2. Do you implement circuit breakers for slow skills?
3. Any built-in OpenClaw profiling tools I'm missing?

I've written up some profiling techniques at https://miaoquai.com/stories/openclaw-skill-performance.html - would love to compare notes!

---

"Every millisecond is a choice between patience and optimization."