All projects

Evidence-first .NET profiling

Performance Skill

Turn performance work into a reproducible evidence trail across managed, native, operating-system, and GPU layers.

StatusActive
TierMaintained
DomainTools
Repositorywieslawsoltes/Performance-Skill

A modular coding-agent skill for cross-platform .NET performance engineering across runtime, memory, latency, startup, production, benchmarking, and GPU/rendering investigations.

Performance Skill treats optimization as an evidence problem. It routes an agent from a measurable symptom to the narrowest appropriate capture, makes command and tool versions explicit, and requires before/after validation in the real application context.

01

Runtime, memory, latency, and startup

02

Benchmarking and production diagnostics

03

macOS, Windows, and Linux toolchains

04

GPU, compositor, and presentation analysis

Where it fits.

  • Coding-agent users investigating .NET performance
  • Application teams building repeatable profiling playbooks
  • Graphics engineers tracing CPU-to-display bottlenecks
macOS

Instruments / xctrace

Windows

WPR, WPA, ETW, WinDbg

Linux

perf, eBPF, procfs

.NET diagnostics

Runtime tool suite

Real output,
not a mockup.

Performance Skill GitHub repository preview
The skill keeps operational procedures, platform guidance, command references, validation, and primary documentation together.

Clear layers.
Replaceable seams.

01

Route

Compact indexes load only the domain and platform guidance required by the investigation.

02

Capture

Version-checked commands collect bounded evidence with workloads and raw artifacts preserved.

03

Analyze

Procedures assign ownership across runtime, native, kernel, I/O, driver, GPU, compositor, and display layers.

04

Validate

Equivalent before/after workloads, tail metrics, variance, and real-application checks prove the result.

8 domainsrouted investigation guides
3 platformsnative profiler lanes
Validatorstructure and command audit
LiveGitHub metadata

From package
to first frame.

Choose the package workflow used by your repository, then start from the smallest working example. Replace VERSION with the current stable or prerelease version shown on NuGet.

01 · Install
Terminal
git clone https://github.com/wieslawsoltes/Performance-Skill.git
# Copy SKILL.md and references into your agent skills directory.
02 · Use · bash
Terminal
$dotnet-performance Profile this application and prove
whether CPU, allocation, I/O, driver, GPU, or presentation
is the dominant bottleneck.

Start focused.
Compose as needed.

The repository is the source of truth.

Use the current documentation and samples, inspect changes, report issues, or support continued open-source maintenance.

Parallels CLI Skill

Search everything.

Loading the index…

move open esc close

Preparing metadata and full-text search…