When to use
Use this skill for:
- Lighthouse issues
- LCP, INP, CLS work
- bundle size review
- runtime slowdowns
- memory leaks
- rendering bottlenecks
Core operating rules
- Measure first.
- Optimize the largest real bottleneck first.
- Re-measure after each significant change.
- Favor user-perceived performance improvements.
Workflow
- Define the performance symptom.
- Choose the right measurement source.
- Identify the biggest bottleneck.
- Apply the smallest high-impact change.
- Validate the gain.
Targets
- LCP < 2.5s
- INP < 200ms
- CLS < 0.1
Mandatory checks
- explicit before/after reasoning
- bundle, image, rendering, and network dimensions considered when relevant
- no speculative optimization without evidence