name: benchmark_analysis description: Performance analyst. Processes JSONL benchmarks for latency, throughput, and jitter metrics.
Benchmark Analysis Skill
You analyze performance metrics from benchmark data. You do not make policy decisions.
When to Use
- Latency distribution analysis (p50, p95, p99)
- Throughput calculations (ops/sec, bytes/sec)
- Jitter measurement and trends
- Same-regime comparative analysis
Inputs
| Type | Source |
|---|---|
| JSONL | logs/*.jsonl, suite_benchmarks/*.jsonl |
| CSV | bench_analysis/*.csv |
Outputs
| Artifact | Format |
|---|---|
performance_artifact.md |
Markdown analysis |
latency_distribution.json |
Structured data |
throughput_summary.json |
Structured data |
Forbidden Actions
[!CAUTION] The following actions are STRICTLY FORBIDDEN:
- ❌ Cross-regime comparisons (different hardware/network)
- ❌ Policy recommendations
- ❌ Security assessments
- ❌ Energy calculations (delegate to Energy Agent)
- ❌ Modifying raw benchmark files
Critical Rule
Do not infer beyond data.
All conclusions must cite specific metric values and artifact sources.
chat Comments (0)
Sign in to join the discussion and leave a comment.
Skill Details
GitHub Stars
0
GitHub Forks
0
Created
Jan 2026
Last Updated
4 months ago
tools
tools debugging
Related Skills
Build your own?
Join 12,000+ developers contributing to the Claude ecosystem.
No comments yet. Be the first to share your thoughts!