BE

benchmark-harness

A specialized skill for automating and documenting CUDA benchmarking processes.

Install

mkdir -p .claude/skills/benchmark-harness && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/12760" && unzip -o skill.zip -d .claude/skills/benchmark-harness && rm skill.zip

Installs to .claude/skills/benchmark-harness

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Use when creating or revising CUDA benchmark runners and result artifacts.
74 chars✓ has a “when” trigger
Intermediate

Key capabilities

  • Define benchmark shapes and dtypes
  • Add warmup and timed iterations
  • Ensure outputs go to artifacts/benchmarks/
  • Include an environment snapshot reference
  • Preserve machine-readable output

How it works

The skill guides the process of defining benchmark parameters, adding warmup and timed iterations, and ensuring the output is saved in a machine-readable format.

Inputs & outputs

You give it
Benchmark parameters like shapes, dtypes, and device tier
You get back
Benchmark results in machine-readable format within artifacts/benchmarks/

When to use benchmark-harness

  • Defining custom benchmark shapes and dtypes
  • Adding warmup iterations to CUDA kernels
  • Recording latency and throughput metrics
  • Generating benchmark artifacts for performance analysis

About this skill

Benchmark Harness

Use this skill when creating or updating benchmark runners.

Workflow

  1. Define benchmark shapes, dtypes, and device tier.
  2. Add warmup and timed iterations.
  3. Ensure outputs go to artifacts/benchmarks/.
  4. Include an environment snapshot reference when the comparison matters.
  5. Preserve machine-readable output.

Minimum Metrics

  • median latency
  • p95 latency
  • throughput when meaningful

When not to use it

  • When the task is not about creating or updating benchmark runners

Limitations

  • It is specific to CUDA benchmark runners
  • It focuses on result artifacts

How it compares

This skill provides a structured workflow for creating and updating CUDA benchmark runners, ensuring consistent output and metric collection, unlike manual ad-hoc benchmarking.

Compared to similar skills

benchmark-harness side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
benchmark-harness (this skill)05moNo flagsIntermediate
google-analytics436moReviewIntermediate
backtesting-trading-strategies1027dReviewIntermediate
spark-optimization42moNo flagsAdvanced

Try saying

Example prompts that trigger this skill in your AI assistant.

You might also like

google-analytics

davila7

Analyze Google Analytics data, review website performance metrics, identify traffic patterns, and suggest data-driven improvements. Use when the user asks about analytics, website metrics, traffic analysis, conversion rates, user behavior, or performance optimization.

43193

backtesting-trading-strategies

jeremylongshore

Backtest crypto and traditional trading strategies against historical data. Calculates performance metrics (Sharpe, Sortino, max drawdown), generates equity curves, and optimizes strategy parameters. Use when user wants to test a trading strategy, validate signals, or compare approaches. Trigger with phrases like "backtest strategy", "test trading strategy", "historical performance", "simulate trades", "optimize parameters", or "validate signals".

1071

spark-optimization

wshobson

Optimize Apache Spark jobs with partitioning, caching, shuffle optimization, and memory tuning. Use when improving Spark performance, debugging slow jobs, or scaling data processing pipelines.

431

ascend-profiling-analysis

Ascend

Analyze Ascend NPU profiling data to identify training performance bottlenecks. Breaks down step-level time into compute, unoverlapped communication, and freetime; within compute, analyzes compute vs memory-bound ratios and cube vs vector utilization to summarize the model's performance bottleneck.

00

spark-optimization

bika11

Optimize Apache Spark jobs with partitioning, caching, shuffle optimization, and memory tuning. Use when improving Spark performance, debugging slow jobs, or scaling data processing pipelines.

00

streamlit

sverzijl

When working with Streamlit web apps, data dashboards, ML/AI app UIs, interactive Python visualizations, or building data science applications with Python

86239

Search skills

Search the agent skills registry