benchmark

A performance testing workflow that measures how quickly a website, its files, and its API requests run before and after a change. It includes page speed, Core Web Vitals (user-focused website speed and stability measures), file sizes, and response times.

In plain words
What is it for?
Use it to set a performance starting point, compare changes, and detect performance regressions before release.
Why use it?
It replaces guesswork with repeatable measurements and helps reveal slowdowns introduced by new code.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/code-saurabh/openskills/benchmark
Any agent
npx skills add CODE-SAURABH/OpenSkills --skill benchmark
Clone the repo
git clone --depth 1 https://github.com/CODE-SAURABH/OpenSkills

Made for: Claude Code, Codex.

Per session 62 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,976 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00062 $0.02976
Opus 5 $0.00031 $0.01488
Sonnet 5 $0.00012 $0.00595
Haiku 4.5 $0.00006 $0.00298

Measured yesterday against content hash 315c24f3990e, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

benchmark scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

The scan reads SKILL.md. This mod also ships 1 executable file (scripts/compare_metrics.py), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

### Quick curl timing
benchmark/SKILL.md · 334 lines

How it starts

The opening of the file, as written. The whole thing — 334 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Benchmark

Benchmarking without a baseline is noise. A benchmark that can't be reproduced is a guess. Every benchmark in this skill produces a number, a method, and a comparison — not an impression.

Performance is a feature. Regressions ship silently. This skill makes regressions visible before they reach users.


Benchmark Principles

  • Measure before you optimize. An optimization without a before measurement is an assumption. Always baseline first.
  • Reproduce exactly. A benchmark is only valid if another engineer can reproduce it with the same setup. Document the environment, the command, and the conditions.
  • Medians lie. Use percentiles. P50 hides the users who are suffering. Always report P75, P95, and P99 alongside the median.
  • Synthetic vs real-user data. Lighthouse in a lab tells you what's possible. RUM (Real User Monitoring) tells you what's happening. Use both.
  • One change at a time. If two things change between benchmarks, you can't attribute the delta to either.
  • A regression is a regression. A 10% slowdown on a fast page is still a regression. Set thresholds and enforce them.

Step 0: Establish the Baseline

Before any optimization or PR review, establish the baseline. Without it, "after" numbers are meaningless.

# Run Lighthouse baseline (3 runs, take median)
npx lighthouse https://your-app.com/key-page \
  --output json \
  --output-path ./baseline.json \
  --chrome-flags="--headless" \
  --throttling-method=simulate \
  --preset=desktop

# Run 3 times and average — single runs are noisy
for i in 1 2 3; do
  npx lighthouse https://your-app.com/key-page \
    --output json \
    --output-path "./baseline-run-$i.json" \
    --chrome-flags="--headless"
done

Record in benchmark-baseline.md:

## Baseline — [Page/Endpoint] — [Date] — [Commit SHA]

| Metric | Run 1 | Run 2 | Run 3 | Median |
|--------|-------|-------|-------|--------|
| LCP    |       |       |       |        |
| INP    |       |       |       |        |
| CLS    |       |       |       |        |
| FCP    |       |       |       |        |
| TTFB   |       |       |       |        |
| Total JS bundle | | | | |
| Total CSS bundle | | | | |
| Network requests | | | | |

**Environment:** macOS M2 / 6-core CPU throttle / Fast 3G simulation
**Tool:** Lighthouse 12.x / Chrome 124
**URL:** https://your-app.com/key-page

Read the full file on GitHub · 334 lines

Files

What ships with it

2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 334 lines · 62 tokens per session scan A 315c24f3990e

Subscribe to this mod's changes

benchmark is a skill published in the GitHub repository CODE-SAURABH/OpenSkills (2 stars, last pushed 1mo ago), licensed MIT. It adds 62 tokens to every session and 2,976 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.