If you're tired of waiting for Instruments to open just to find performance bottlenecks, this gets you Time Profiler data entirely from the command line. You attach to or launch your app with xctrace, extract the samples to XML, then symbolicate and rank the hotspots using atos and the runtime load address from vmmap. It includes Python scripts that do the extraction and ranking for you. The workflow is solid if you're already living in the terminal and just want stack traces with line numbers pointing to your top 30 CPU hogs. One catch: you need to match your binary exactly and grab the ASLR offset while the app is running, but the docs walk you through it.
npx skills add https://github.com/steipete/agent-scripts --skill native-app-performance