What problem does it solve? Android apps often feel slow or stuttery without developers knowing why. This Skill measures UX-relevant performance (app startup time, frame jank, loading-state correctness) on a physical device against fixed budgets, then guides interactive remediation with re-measurement to prove each fix worked. ## Core Features & Use Cases - MEASURE phase: Runs Macrobenchmark TTID/TTFD startup metrics, FrameTimingMetric jank percentiles, Perfetto trace analysis, and loading-state audits against response-time thresholds, with full run-condition capture (device, build type, refresh rate, thermal state). - FIX phase: Applies one remediation at a time behind an operator approval gate — Baseline Profiles, deferred initialization, moving work off the main thread, corrected loading-state UI — then rebuilds and re-measures to attribute the delta. - Resumable runs: Checkpoints baselines and approval decisions to .resume/ files so multi-session measurement campaigns can resume safely. - Use Case: Your app's cold start feels slow. The Skill captures a Perfetto trace, names the cost centre (e.g., blocking I/O in Application.onCreate()), proposes the fix, and re-measures to confirm TTID dropped under the 500 ms budget. ## Quick Start Ask the assistant to measure your app's cold startup time and fix any jank on the main screen using the android-perceived-performance skill.
