What problem does it solve?
Diagnosing CPU regressions in the Wox launcher is hard because Go pprof data, native renderer activity, and background work after hide can each distort the picture. This Skill provides a deterministic, repeatable workflow to measure query-active and hidden CPU in the real single-process Wox debug build and attribute hotspots to the correct owner.
Core Features & Use Cases
- Deterministic workload replay: Drives the real Wox debug build through wox_automation with seeded mixed or fixed queries, keeping warm-up, profiling, and sampling in separate identical runs.
- Cross-platform CPU sampling: Captures Go runtime/pprof profiles, macOS top/sample process data, and Windows TotalProcessorTime deltas or ETW stacks, with rules for handling Windows blocked-call profiling artifacts.
- Memory guardrails: Requires before/after process footprint and Go retained-heap comparisons so CPU optimizations never trade for unbounded caches or retained native resources.
- Use Case: When Wox shows high idle CPU after being hidden, run the hidden workload, sample the settled process for 30 seconds, capture native stacks, and classify whether a recurring timer, animation, or renderer stack is a real regression.
Quick Start
Use the wox-cpu-debug skill to measure Wox's hidden CPU usage on this machine and tell me whether any background work continues after the launcher is hidden.