Przejdź do treści
Biblioteka skilli
DevOps i chmura · Planowanie

measure-everything

/measure-everything

Ustala, jak mierzyć przed budową: ślady, metryki, logi, Web Vitals i czas CI, zanim napisze się własny kod pomiarowy.

Repozytorium
skills ↗
Licencja
AGPL-3.0
Ostatnia zmiana repozytorium
13.11.2026

Ocena i bezpieczeństwo

90/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść25 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
4
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
11
Przestarzałość

Brak sygnałów. Repozytorium zmienione 13.11.2026, bez odwołań do wycofanych modeli i API.

Licencja

AGPL-3.0, copyleft: przeróbki udostępniane innym muszą mieć tę samą licencję.

Co robi i kiedy

Ustala, jak mierzyć przed budową: ślady, metryki, logi, Web Vitals i czas CI, zanim napisze się własny kod pomiarowy.

Oryginalny opis (EN)

Decide how to measure before building or fixing something. Use when adding traces, metrics or logs, checking that telemetry really leaves the process, measuring Web Vitals or page transitions, timing CI, or when about to write custom measurement code.

Jak działa

  1. Why this skill exists and the state it aims for
  2. Rules
  3. Which means to use
Sekcje z nagłówków SKILL.md, w oryginale · 108 linii, 665 słów
Podgląd SKILL.md
---
name: measure-everything
description: Decide how to measure before building or fixing something. Use when adding traces, metrics or logs, checking that telemetry really leaves the process, measuring Web Vitals or page transitions, timing CI, or when about to write custom measurement code.
---

# measure-everything

## Why this skill exists and the state it aims for

This skill exists because an agent cannot see what its code does in production.
It fills the gap with guesses, and in the next step a guess looks the same as a
fact. Agents also tend to write their own timers, wrappers and log lines. Those
measure only what the author thought of. They drift from the code, and they
never reach the tools that people already read.

The state this skill aims for is that every decision rests on a measurement.
The measurement is taken where the software really runs. The official features
and auto-instrumentation of the tools in use collect it. So the measurement
keeps up with the code by itself.

## Rules

Measure before you implement or fix. First decide which measurement would
settle the question. Then run it, and judge only by its result.

Before you build a feature, make sure you can debug it yourself. You need a way
to reproduce it, logs or spans, and a path to read them.

Measure in production, or in a staging environment proven to match it. Numbers
from a local machine or a test alone are not grounds for a decision.

Make everything that spends time or money a span. That covers outbound requests,
model calls, child processes and file I/O. Put the span in the code that spends
the time. Do not wrap a command or a process only to time it.

Pick the means in this order.

1. The official feature or auto-instrumentation of the platform or library.
2. An existing tool that already does it.
3. A `@r_masseater/*` measurement package from masseater.dev, which only fills
   what the first two miss.
4. Your own code. Write in the commit message what you searched for and why it
   did not fit.

Configure exporters with the standard `OTEL_*` environment variables. Send only
when they are set, and never hard-code an endpoint.

When a test can check a measurement, write that test. Auto-instrumentation that
silently stops exporting is a regression.

## Which means to use

For traces, metrics and logs of a Node.js process, use
[OpenTelemetry zero-code
instrumentation](https://opentelemetry.io/docs/zero-code/js/).
It is the `@opentelemetry/auto-instrumentations-node/register` preload,
configured with `OTEL_*`.

In Effect programs, use `Effect.fn` and `Effect.withSpan`. Export them with
`OtlpTracer.layerFromConfig` from `effect/observability`.

On Cloudflare, use the measurement the platform already has. Workers tracing,
Workers logs and D1 query analytics cover most needs. Read
[cloudflare.md](references/cloudflare.md) before you add any of them.

For the duration of a Vitest run, turn on
[`experimental.openTelemetry`](https://vitest.dev/guide/open-telemetry). Each
test becomes a span, and spans created in a test join its trace.

For Playwright tests, use
[Playwright
OpenTelemetry](https://endform.dev/blog/introducing-playwright-opentelemetry).
Tests and their steps become spans, and the trace context reaches the app
through the requests of the browser.

For model calls through OpenRouter, turn on
[Broadcast](https://openrouter.ai/docs/guides/features/broadcast/otel-collector)
in its settings. It sends cost, tokens and latency of each call over OTLP
without code changes.

For Claude Code sessions, set `CLAUDE_CODE_ENABLE_TELEMETRY=1` with the usual
`OTEL_*` variables. It exports metrics and events, as the
[monitoring guide](https://code.claude.com/docs/en/monitoring-usage) describes.
Traces are in beta and also need `CLAUDE_CODE_ENHANCED_TELEMETRY_BETA=1`.

On Fly.io, read the
[built-in metrics](https://fly.io/docs/monitoring/metrics/) from the managed
Prometheus. The managed Grafana already has dashboards for them.

To check that telemetry follows the semantic conventions, run the live-check of
[OpenTelemetry Weaver](https://opentelemetry.io/blog/2025/otel-weaver/) in CI.
It exits with 1 when it finds a violation. Tests that export telemetry can run
the same check. Pass `semconv: true` to `otlp()` of `@r_masseater/vitest-otlp`.

For the size of a published package, use
[size-limit](https://github.com/ai/size-limit). Set the limit to the current
size, so that any growth fails and its reason goes into the commit message.

To assert in a test that telemetry was really exported, use
[`@r_masseater/vitest-otlp`](https://github.com/masseater/masseater.dev/tree/main/libs/vitest-otlp).

To test the automatic spans of a Worker, run it in Miniflare with
`tailWorker()` of `@r_masseater/vitest-otlp/workers`.

For Web Vitals of page loads and same-origin link transitions, use
[`@r_masseater/web-vitals-flow`](https://github.com/masseater/masseater.dev/tree/main/libs/web-vitals-flow).
Read [web-vitals.md](references/web-vitals.md) before you run it.

For durations of GitHub Actions workflows, jobs and steps, use
[otel-cicd-action](https://github.com/corentinmusard/otel-cicd-action). Read
[ci.md](references/ci.md) before you add it.
Pliki w paczce (4)
  • SKILL.md 5 KB
  • references/ci.md 1 KB
  • references/cloudflare.md 3 KB
  • references/web-vitals.md 2 KB

Wersje

W bazie nie ma innych skilli o nazwie measure-everything.

Migawka bazy z 13.11.2026 · wersja paczki bcc5ed6221a6 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 100analyze-local-telemetry
    Analizuje lokalnie przechwycone ślady, logi i metryki OTel: przegląda spany, szuka wolnych operacji, koreluje logi ze śladami, odpytując dev/local-telemetry przez SQL po HTTP.
  • 100observability-monitoring
    Projektuje i przegląda logi, metryki, ślady, alerty, SLO, dashboardy i zdarzenia audytowe dla aplikacji webowych, API, SaaS, backendów mobilnych i systemów AI.
  • 100observability
    Dodaje lub zmienia instrumentację obserwowalności: logowanie strukturalne, ślady i spany OpenTelemetry, metryki RED, dashboardy, alerty, kontrakty pól telemetrii i runbooki monitoringu.
  • 100cx-telemetry-querying
    Odpowiada na pytania o dane telemetryczne Coralogix: badanie incydentów, debugowanie, błędy, logi, trasy, spany i opóźnienia żądań.
  • 90datadog
    Analizuje logi, metryki, ślady i incydenty w Datadog dla projektu Speakeasy, gdy pojawiają się błędy lub opóźnienia.
  • 100observability-advisor
    Projektuje i przegląda logi, metryki, ślady, SLO i alerty dla niezawodnych systemów oraz wskazuje luki w pokryciu telemetrią.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.