hyperfine-benchmarking

v2026.09.24

Use this skill to benchmark shell commands reliably with warmup runs, statistical summaries, and exportable artifacts using hyperfine.

GitHub
安装命令
npx skhub add akillness/hyperfine-benchmarking
Markdown
SKILL.md

hyperfine-benchmarking

Benchmark CLI commands with reproducible methodology instead of one-off time output.

When to use this skill

  • Compare two or more command variants (tool-a vs tool-b)
  • Validate performance impact before/after script changes
  • Attach benchmark evidence to PRs

Instructions

  1. Confirm tool availability.
  2. Keep input/workdir/environment stable across compared commands.
  3. Run warmups and enough runs for stable variance.
  4. Export JSON/markdown outputs for review.
  5. Summarize relative speedup + risk notes.

Examples

Availability check

hyperfine --version

Two-command comparison

hyperfine \
  --warmup 3 \
  --min-runs 10 \
  'cmd_a --with flags' \
  'cmd_b --with flags'

Parameter sweep

hyperfine \
  --warmup 3 \
  --parameter-list mode fast,balanced,thorough \
  'mytool --mode {mode} input.txt'

Export artifacts

hyperfine \
  --warmup 3 \
  --min-runs 10 \
  --export-json benchmark.json \
  --export-markdown benchmark.md \
  'cmd_a' 'cmd_b'

Best practices

  • Prefer relative speedup and confidence ranges over single-run claims.
  • Do not compare commands with different semantics unless outputs are normalized.
  • If variance is high, increase runs or reduce background noise before concluding.
  • Record dataset/path and exact command strings in PR text.

References

发现
标签

此技能尚未发布标签。

版本
最新版本元数据

版本

v2026.09.24

发布时间

Sep 24, 2026

分类

未分类

许可证

未指定

源路径

.agent-skills/hyperfine-benchmarking

默认分支

main

最新提交

f579bfe

Tree SHA

34a09b3