Навыки measure-dont-guess

📦

measure-dont-guess

Name: measure-dont-guess
Author: masseater

Ревизия содержимого r1 Средний риск ⚙️ Внешние команды

Measure Evidence Before Changing Code

Assumptions can lead to ineffective fixes and misleading performance claims. This skill guides assistants through hypotheses, predefined criteria, repeatable measurements, and evidence-based decisions.

Поддерживает: Claude Codex Code(CC)

📊 67 Адекватно

Maintain this Skill? Claim it 📊 67 Адекватно

Установить с помощью моего Агента

Скопируйте этот запрос в своего Агента. Он содержит каноническую страницу Skill и манифест.

Запрос агента

Review the Skillstore skill "measure-dont-guess" from https://skillstore.io/skills/masseater-measure-dont-guess.md and its manifest at https://skillstore.io/api/skills/masseater-measure-dont-guess/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.

Ваш Агент по-прежнему должен показать план и запросить все подтверждения, требуемые политикой безопасности.

Ресурсы для AI-агентов

Используйте эти ссылки, когда AI-агенту, crawler или script нужен чистый контекст вместо полной страницы.

Markdown-детали GET /skills/masseater-measure-dont-guess.md Подписанный manifest GET /api/skills/masseater-measure-dont-guess/manifest Подписанный lockfile GET /api/skills/masseater-measure-dont-guess/lockfile

Протестировать

Использование «measure-dont-guess». Illustrative request: Investigate intermittent checkout failures without assuming the database is responsible.

Ожидаемый результат:

Question: At which boundary does the checkout request first become invalid?
Hypotheses: request validation, service transformation, or database handling.
Observation plan: Record redacted boundary values and correlation identifiers in an approved environment.
Decision criterion: Identify the earliest boundary where valid input produces invalid output.
Status: Root cause remains unconfirmed until reproduction and measurements are available.

Использование «measure-dont-guess». Illustrative request: Determine whether caching improves a slow endpoint.

Ожидаемый результат:

Hypothesis: Repeated reads account for most endpoint latency.
Baseline plan: Measure endpoint and read spans under the same representative workload across repeated runs.
Decision criterion: Agree on the required latency reduction and acceptable error rate before testing.
Experiment: Change only caching behavior, repeat measurements, and inspect other operations for regressions.
Status: No performance improvement is claimed before measurements.

Аудит безопасности

Средний риск

v1 • 24.10.2026 Открыть версионный отчет

All seven static alerts are false positives: Markdown link labels are not shell execution, and the entropy alerts concern readable Japanese documentation. No evidence found of prompt injection, obfuscated payloads, or intentional data exfiltration. A separate confidentiality risk arises from recommending production diagnostics and logging boundary values without redaction guidance.

Просканировано файлов

133

Проанализировано строк

Пункты проверки

Ложные срабатывания проигнорированы

Подтверждённые проблемы безопасности (1)

Средний

SKILL.md:21-27 SKILL.md:32 usecases/find-root-cause.md:8

Production Diagnostics Lack Sensitive Data Safeguards

The workflow prioritizes production measurements and instructs logging incoming and outgoing values at layer boundaries. Without redaction, credentials or personal data could enter logs. Recording measurement results in commit messages could further persist sensitive diagnostic content.

Production measurement and boundary-value logging are explicit instructions, and the reviewed workflow provides no redaction guidance. Actual disclosure depends on diagnostic data and execution.

Факторы риска

⚙️ Внешние команды (4)

SKILL.md:39 SKILL.md:40 SKILL.md:41 SKILL.md:45

Аудитор:: codex

Поделиться и цитировать этот отчет

Делитесь версионным отчетом об оценке, нейтральным значком, встраиваемой карточкой и цитатами. Skillstore публикует доказательства, не решая, безопасен ли этот Skill.

Открыть версионный отчет

Копировать ссылку на отчёт

https://skillstore.io/skills/masseater-measure-dont-guess/audits/1?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Значок Markdown

[![Skillstore security assessment](https://skillstore.io/badges/skills/masseater-measure-dont-guess/security.svg)](https://skillstore.io/skills/masseater-measure-dont-guess?utm_source=security_passport_badge)

Значок HTML

<a href="https://skillstore.io/skills/masseater-measure-dont-guess?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/masseater-measure-dont-guess/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Встраиваемая карточка

<iframe src="https://skillstore.io/embed/skills/masseater-measure-dont-guess.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

Академические ссылки (APA · BibTeX · CFF)

Цитата APA

masseater. (2026). measure-dont-guess security audit report (audit version 1) [Author version unspecified]. Skillstore. https://skillstore.io/skills/masseater-measure-dont-guess/audits/1

Цитата BibTeX

@techreport{masseater-masseater-measure-dont-guess-2026,
  author = {masseater},
  title = {measure-dont-guess security audit report (audit version 1)},
  institution = {Skillstore},
  year = {2026},
  number = {1},
  url = {https://skillstore.io/skills/masseater-measure-dont-guess/audits/1},
  note = {Author version unspecified}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "measure-dont-guess security audit report (audit version 1)"
version: "unspecified"
type: report
authors:
  - name: "masseater"
date-released: "2026-10-24"
url: "https://skillstore.io/skills/masseater-measure-dont-guess/audits/1"
identifiers:
  - type: other
    value: "skillstore:masseater-measure-dont-guess:audit:1"
    description: "Skillstore immutable audit report identifier"

Оценка Skillstore

Почему такая оценка Достоверность доказательств: Средний

Архитектура

Сопровождаемость

Контент

Сообщество

Соответствие спецификации

Что вы можете построить

Investigate a Reproducible Bug

Trace a failing request across application layers, isolate the source of invalid values, and verify one targeted fix.

Verify a Performance Improvement

Collect a representative baseline, change one measured bottleneck, and compare repeated results while checking for regressions.

Define Feature Acceptance Evidence

Turn feature requirements into failing acceptance checks before implementation, then verify the completed behavior with the same checks.

Попробуйте эти промпты

Beginner: Define a Measurement

Use measure-dont-guess for [question]. List hypotheses, the measurement needed for each, and decision criteria before proposing code changes.

Intermediate: Find the Root Cause

Investigate [bug] using [reproduction and logs]. Review recent changes, propose redacted boundary observations, and define a failing test before one targeted fix.

Advanced: Compare Performance

Evaluate [proposed optimization] in [approved representative environment]. Define baseline spans, repeat measurements, report variation, and check regressions after one change.

Advanced: Validate Across Layers

Plan validation for [feature] across [layers]. Define failing acceptance tests, redacted observations, and decision criteria. Identify missing access before any production action.

Лучшие практики

Write hypotheses and acceptance criteria before collecting results, then keep experiment conditions consistent.
Use approved representative environments, redact sensitive diagnostics, and sanitize evidence before adding it to commits.
Change one factor at a time, repeat measurements, and report uncertainty or missing access explicitly.

Избегать

Treating an unmeasured explanation as a confirmed root cause.
Changing acceptance criteria after seeing results or combining unrelated fixes.
Logging raw production inputs or committing diagnostic results without checking for sensitive data.

Часто задаваемые вопросы

What does this skill provide?

It provides measurement-first instructions for debugging, performance comparisons, and feature validation. It does not include an executable measurement engine.

Can I use it with Claude, Codex, and Claude Code?

The workflows can guide these assistants. Actual measurements depend on the tools, permissions, and environment available to each assistant.

Does it require the referenced TypeScript template?

No. The general workflow applies elsewhere, but Effect and Cloudflare instructions must be mapped to equivalent tools in your repository.

Must measurements run in production?

The source prioritizes production and otherwise calls for matching staging conditions. Obtain authorization and define diagnostic safeguards before accessing either environment.

What happens when measurements are unavailable?

The workflow directs the assistant to explain what is missing and label unmeasured claims as speculation instead of proceeding on assumptions.

How should logs and measurement records be handled?

Redact secrets and personal data, restrict access, and sanitize results before committing them. These safeguards must supplement the source instructions.