linux-perf

What it does

This skill enables AI agents to interact with the Linux Performance Monitoring Framework (perf) to analyze system performance data. It allows for the execution of perf commands to profile applications, identify bottlenecks, and inspect hardware events directly from within an agent environment.

When to use it

• Profiling a specific application binary to understand CPU usage patterns and hotspots.
• Investigating system-wide latency issues by analyzing kernel tracepoints and hardware counters.
• Generating performance reports for debugging slow-running services on Linux infrastructure.
• Verifying the impact of code optimizations by comparing perf data before and after changes.

Key capabilities

• Execute perf commands to record and analyze performance metrics.
• Profile user-space applications to identify inefficient functions or threads.
• Inspect hardware events and kernel-level performance counters.
• Generate output suitable for further analysis or reporting within the agent workflow.

Example prompts

• "Run a perf profile on /usr/bin/myapp with 10 seconds of recording time and display the top 5 functions by CPU usage."
• "Analyze system performance using perf top to identify the most active kernel functions over the last minute."
• "Generate a report showing hardware cache misses for the process with PID 1234 using perf stat."

Tips & gotchas

Ensure the target Linux environment has the linux-perf toolchain installed and that the user agent has sufficient permissions to access performance data. Be aware that running perf on production systems may incur slight overhead, so it is best suited for staging or development environments unless configured carefully.