What problem does it solve?
When users report that a browser extension makes a page leak memory, freeze, or burn CPU, it is hard to tell whether the extension or the site itself is at fault. This Skill turns vague performance complaints into measured, attributed root causes with reproducible pass/fail numbers.
Core Features & Use Cases
- Attribution ladder experiments: Run controlled comparisons (no-extension baseline, extension-on, static-page control, feature-isolation) to separate extension faults from the site's own runaway loops.
- Metric-driven failure signatures: Interpret CDP metrics (Nodes, JSHeapUsedSize, TaskDuration, layout counters) to distinguish real leaks from harmless churn and detached-DOM CPU storms.
- Starved-renderer profiling: Capture CPU profiles through the browser-process Tracing domain when the renderer main thread is too busy to answer normal CDP probes, then map minified frames back to site or extension code.
- Use Case: A user reports the extension freezes a news site. You run the Puppeteer harness, discover CDP Nodes exploding while attached DOM stays flat, trace 90% of CPU to the site's own jQuery overflow loop, and ship an exclusion rule instead of rewriting extension code.
Quick Start
Use the extension-perf-forensics skill to diagnose why the extension makes this reported page freeze and produce an attributed root-cause report with pass/fail metrics.