Skip to content

Add profiling-dotnet-apps skill (+11.0% eval improvement) - #80

Closed
mrsharm wants to merge 2 commits into
dotnet:mainfrom
mrsharm:musharm/profiling-dotnet-apps-skill
Closed

Add profiling-dotnet-apps skill (+11.0% eval improvement)#80
mrsharm wants to merge 2 commits into
dotnet:mainfrom
mrsharm:musharm/profiling-dotnet-apps-skill

Conversation

@mrsharm

@mrsharm mrsharm commented Feb 23, 2026

Copy link
Copy Markdown
Member

Summary

Adds the profiling-dotnet-apps skill that teaches .NET diagnostic CLI tool workflows.

Eval Results

Metric Score
Overall Improvement +11.0%
Threshold 10%
Result PASS

What the Skill Teaches

  • dotnet-counters for real-time performance monitoring (specific counter names and arguments)
  • dotnet-trace for CPU profiling with provider keywords
  • dotnet-dump for crash/hang analysis
  • dotnet-gcdump for lightweight GC heap snapshots
  • PerfView for trace analysis

Why This Skill Passes

The model generates approximate CLI tool commands but gets specific argument syntax and provider keywords wrong. This skill provides the exact command formats and provider GUIDs needed.

Files

  • src/dotnet/skills/profiling-dotnet-apps/SKILL.md
  • src/dotnet/tests/profiling-dotnet-apps/eval.yaml

@ViktorHofer

ViktorHofer commented Feb 23, 2026

Copy link
Copy Markdown
Member

Please share the results as a comment similar to #75 (comment)
Also make sure that you test at least 3 runs (--runs 3 for local validation): https://github.com/dotnet/skills/pull/82/changes

Teaches .NET diagnostic CLI tool workflows: dotnet-counters for real-time
monitoring, dotnet-trace for CPU profiling, dotnet-dump for crash analysis,
and dotnet-gcdump for memory snapshots.

Eval results: +11.0% improvement over baseline (threshold: 10%)
Includes eval.yaml with profiling scenario + negative test.
@mrsharm
mrsharm force-pushed the musharm/profiling-dotnet-apps-skill branch from 9814606 to cebe1c1 Compare February 23, 2026 14:52
@mrsharm mrsharm changed the title Add profiling-dotnet-apps skill with eval Add profiling-dotnet-apps skill (+11.0% eval improvement) Feb 23, 2026
@mrsharm

mrsharm commented Feb 25, 2026

Copy link
Copy Markdown
Member Author

Skill Validation Results — profiling-dotnet-apps

Skill Test Baseline With Skill Δ Verdict
profiling-dotnet-apps Profile slow API response times 4.7/5 5.0/5 +0.3

Overall improvement: -19.6% (3 runs)

The quality score improved (4.7 → 5.0), but the pairwise judge penalizes token/time overhead, resulting in a net negative composite score. The pairwise judge infrastructure also experienced intermittent timeouts across multiple runs, contributing to score instability.

Model: claude-opus-4.6 | Judge: claude-opus-4.6

@mrsharm

mrsharm commented Feb 25, 2026

Copy link
Copy Markdown
Member Author

Closing since this is a futile experiment.

@mrsharm mrsharm closed this Feb 25, 2026
moesac0970 pushed a commit to moesac0970/skills that referenced this pull request Jul 4, 2026
… answer stated up front, MTP detection removed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants