Profile macOS applications with Instruments traces to analyze CPU, GPU, shader, and memory behavior.
npx skills add https://github.com/Kr1sso/xtrace-skill --skill instruments-kr1sso
Please help me install this Agent Skill. Skill: instruments Source: https://github.com/Kr1sso/xtrace-skill Command: npx skills add https://github.com/Kr1sso/xtrace-skill --skill instruments-kr1sso
xtrace, then analyze flat summaries, timelines, call trees, and before/after diffs with trace-analyze.py. - GPU & Shader Analysis: Capture Metal System Trace recordings, inspect GPU state residency, command buffers, shader hotspots, and generate shader flamegraphs or speedscope views. - Memory Analysis: Detect leaks, track memory growth over time, and inspect heap allocations with trace-memory.py. - Use Case: A developer builds a Metal app, runs xtrace --gpu --shader-timeline ./my_app, and immediately gets GPU utilization summaries plus interactive shader stack exploration in speedscope. ## Quick Start Profile my application by running xtrace on the built binary and show me the top CPU hotspots and a flamegraph.High-intent search queries and answers about installing and using this skill.
Prefix any command with xtrace, for example `xtrace -d 10 ./my_app`, to record a Time Profiler trace and print a CPU summary. Then run trace-analyze.py summary, timeline, or calltree on the trace for deeper analysis.
Run trace-flamegraph.sh on the .trace file to produce an SVG flamegraph, or pipe the trace into trace-speedscope.sh for interactive analysis. It uses inferno or flamegraph.pl when installed, with a built-in SVG fallback.
Yes, use `xtrace --gpu` for Metal System Trace or add `--shader-timeline` to enable shader-profiler rows. Then trace-gpu.py reports GPU state and command buffers, while trace-shader.py extracts shader hotspots and callsites.
xtrace works on both Apple Silicon and Intel Macs for standard templates like Time Profiler and Metal System Trace. Only the Processor Trace template requires Apple Silicon with Developer Tools enabled in System Settings.
Missing function names mean the binary lacks debug symbols. Rebuild with symbols: `-g -O2` for C/C++, `swift build -c release -Xswiftc -g` for Swift, or `CARGO_PROFILE_RELEASE_DEBUG=true` for Rust.
No, xctrace and Instruments CLI only record .trace files. A .gputrace bundle requires host-app code calling MTLCaptureManager or capture via the Xcode Metal Debugger; trace-gputrace.py can then inspect the bundle.
web-infra-devTrace Rust function calls and convert them to perfetto protobuf traces.
softsparkProfile Python, Node.js, SQL, and frontend code to identify performance bottlenecks.
rshankrasDiagnose Apple app performance issues with Instruments profiling and SwiftUI rendering guidance.
robinebersAutomates building, running, and debugging macOS apps with Xcode and Swift workflows.