What problem does it solve? Web performance work fails when teams misconfigure Lighthouse CI, set arbitrary score budgets that block deploys, or misread the gap between lab scores and CrUX field data (especially the TBT vs INP divergence). This Skill encodes the correct Lighthouse 12.6.1 / LHCI 0.15.x patterns so audits are reproducible, budgets are grounded in baselines, and field data is interpreted correctly. ## Core Features & Use Cases - LHCI CI Setup: Production-ready lighthouserc configuration with collect/assert/upload blocks, GitHub Actions workflows, and the critical fetch-depth: 20 gotcha for ancestor detection. - Performance Budgets: Guidance on minScore vs maxNumericValue assertions, median-run vs pessimistic aggregation, and the measure-baseline-first workflow before gating CI. - Lab vs Field Reconciliation: A framework for diagnosing why a Lighthouse score of 90 can coexist with failing CrUX INP, using the PSI API's loadingExperience and lighthouseResult blocks. - Use Case: A developer sees "Lighthouse says 92 but PageSpeed Insights says INP is Poor." The Skill walks through pulling the PSI API response, identifying the TBT/INP gap, and locating heavy interaction handlers in a Performance trace. ## Quick Start Ask the AI to set up Lighthouse CI with a performance budget for your Next.js app using this skill's starter template and configuration guides.
