agentleFS
woodbearywoodbeary/later-dog/.claude/skills/benchmark-checklist/SKILL.md

benchmark-checklist

Skill2 starsChanged 29 days agoApache-2.0View on GitHub

Check any measured number before it is reported: label what kind of number it is, make the runs fair, clean and repeated, confirm the timer covered the real work, test the result against what limits it, and show it matters end to end.

54 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use ketchup-video, principle-explain-the-number and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • shiwenbin1617Skill

    ketchup-video

    Catch the user up like /ketchup, with a short subtitled video of slides for what happened and the waiting calls as text under it. Use for /ketchup-video.

    Also in 20 projects that use this

  • shiwenbin1617Skill

    principle-explain-the-number

    Apply before you measure, and before you trust, report, or act on a number you measured: a speedup, a regression, a throughput, a latency, or an eval result. Find what limits it, and rule out that it measured something other than the work you think.

    Also in 35 projects that use this

  • drafaelSkill

    in-your-own-words

    Restate the user's goals and the problem they are trying to solve to check understanding. Manual invocation only; use when the user explicitly invokes in-your-own-words or requests this understanding check. Do not solve the task or rewrite the assistant's last answer.

    Also in 33 projects that use this

  • instructureSkill

    ketchup

    Rebuild context after /clear by analyzing uncommitted changes across Canvas LMS and all plugin repositories

    Also in 32 projects that use this

  • jaydubya818Skill

    principle-test-behavior-not-implementation

    Apply when tests mirror internals instead of asserting observable outcomes.

    Also in 37 projects that use this

  • jaydubya818Skill

    principle-attack-the-premise

    Apply when repeated fixes fail and a shared assumption may be wrong.

    Also in 36 projects that use this

  • +2 more from shiwenbin1617
  • +3 more from jaydubya818

The file

Files
SKILL.md

Benchmark checklist

Use before a speed, latency, throughput, cost, capacity or count goes into a PR, a doc, a job observation or a reply. A number goes out with its receipt: the commands, the raw runs and a verdict. Use the later.dog MCP server. Discover tools by the names below; client-specific prefixes may vary.

1. Label it

  • Measured: it passed everything below.
  • Ballpark: one run, called a ballpark. It still needs clean output and a checked timer (steps 3 and 4).
  • Target: a goal, such as a monthly PR count or a concurrency level. Never write it as an outcome.
  • Unknown: there is no source. Say unknown instead of estimating, the way record_job_observation leaves a missing cost unknown.

2. Set up

Write the sentence you expect to report before you measure, and read the code that produces the number. Note the machine: model, cores, what else is running. Give every variant its real configuration (the same data and warm-up, each with its own best settings) so the old path is not left untuned to flatter the new one.

3. Run

Five or more runs per variant, alternating between variants. Keep every raw result. A run with an error, a retry, a timeout or a skipped step is a failed run: report it, never average it in.

4. Check

  • The timer. The timed region holds the work you describe, and that work happened: output produced, rows written, request answered. An early return or an empty input makes anything fast.
  • The limit. Name what stops the number from being much better (CPU, disk, network, a provider's rate limit, the supervisor's concurrency limit) and do the arithmetic. A result beyond what the hardware or a contract allows means the measurement is wrong.
  • The person. Show the difference where someone using later.dog, or a cloud job, would notice it. A microbenchmark win alone is not a product claim.

5. Report

Median and range per variant, and a verdict: better, worse, no difference beyond the noise, or inconclusive. Put the commands, raw numbers and machine notes beside the claim: in the PR, on a page under docs/verification/ (as persistence-performance.md does), or with record_job_observation for a cloud job. For cloud work, read outcomes from laterdog_workspace and keep PRs opened, verified and merged as three separate numbers.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.