regression-test
Detect LLM Prompt Regressions with Promptfoo
Prompt changes can silently reduce output quality. This skill builds golden datasets, runs Promptfoo evaluations, and summarizes failures and recent trends.
Установить с помощью моего Агента
Скопируйте этот запрос в своего Агента. Он содержит каноническую страницу Skill и манифест.
Review the Skillstore skill "regression-test" from https://skillstore.io/skills/7alexhale5-rgb-regression-test.md and its manifest at https://skillstore.io/api/skills/7alexhale5-rgb-regression-test/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.
Протестировать
Использование «regression-test». Initialize regression testing for this prompt project.
Ожидаемый результат:
Initialized the Promptfoo configuration and golden dataset directory. The results directory is ignored, and the suite is ready for its first case.
Использование «regression-test». Run the golden tests after my system prompt change.
Ожидаемый результат:
Regression results: 12 total, 11 passed, and 1 failed. The auth-flow case failed its expected behavior rubric and needs review.
Использование «regression-test». Show recent prompt quality trends.
Ожидаемый результат:
Recent trend: four of five runs passed completely. One auth-flow failure appeared in the latest run, so quality is declining.
Аудит безопасности
БезопасноAll 38 static findings are false positives caused by Markdown syntax, bounded shell examples, or a Promptfoo message template. No malicious intent, prompt injection, credential handling, or unauthorized network behavior was found.
Факторы риска
⚙️ Внешние команды (35)
🌐 Доступ к сети (1)
📁 Доступ к файловой системе (1)
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.
Копировать ссылку на отчёт
https://skillstore.io/skills/7alexhale5-rgb-regression-test/audits/1?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportЗначок Markdown
[](https://skillstore.io/skills/7alexhale5-rgb-regression-test?utm_source=security_passport_badge)Значок HTML
<a href="https://skillstore.io/skills/7alexhale5-rgb-regression-test?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/7alexhale5-rgb-regression-test/security.svg" alt="Skillstore security assessment" loading="lazy"></a>Встраиваемая карточка
<iframe src="https://skillstore.io/embed/skills/7alexhale5-rgb-regression-test.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
7alexhale5-rgb. (2026). regression-test security audit report (audit version 1) [Author version unspecified]. Skillstore. https://skillstore.io/skills/7alexhale5-rgb-regression-test/audits/1Цитата BibTeX
@techreport{7alexhale5-rgb-7alexhale5-rgb-regression-test-2026,
author = {7alexhale5-rgb},
title = {regression-test security audit report (audit version 1)},
institution = {Skillstore},
year = {2026},
number = {1},
url = {https://skillstore.io/skills/7alexhale5-rgb-regression-test/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: "regression-test security audit report (audit version 1)"
version: "unspecified"
type: report
authors:
- name: "7alexhale5-rgb"
date-released: "2026-08-21"
url: "https://skillstore.io/skills/7alexhale5-rgb-regression-test/audits/1"
identifiers:
- type: other
value: "skillstore:7alexhale5-rgb-regression-test:audit:1"
description: "Skillstore immutable audit report identifier"
Оценка Skillstore
Почему такая оценка Достоверность доказательств: СреднийЧто вы можете построить
Validate prompt changes
Compare updated prompts against recorded expectations before merging code changes.
Check model migrations
Rerun a stable golden dataset after changing a model or provider version.
Enforce release quality
Run prompt evaluations in continuous integration and stop releases when assertions fail.
Попробуйте эти промпты
Initialize Promptfoo regression testing in this repository. Show the files created and explain any prerequisites that are missing.
Add a golden test named {name}. Ask for the system prompt, user input, expected behavior, and optional variables before writing it.Run all golden prompt tests. Summarize the pass rate, duration, and each failed assertion with expected and actual behavior.
Review the last five evaluations, identify new failures, compare pass-rate trends, and recommend quality gates for {release_workflow}.Лучшие практики
- Review every golden case before committing it, and exclude prompts containing secrets or sensitive data.
- Describe expected behavior precisely enough that rubric assertions produce consistent results.
- Run the suite after prompt or model changes and before every release.
Избегать
- Do not treat one current model response as a permanent definition of correct behavior.
- Do not commit API keys, private prompts, personal data, or confidential outputs in golden files.
- Do not ignore unstable rubric failures without reviewing the assertion and provider behavior.
Часто задаваемые вопросы
What must be installed?
Which files does the skill create?
Can these tests run in continuous integration?
Should sensitive prompts be stored as golden cases?
Can I use a different model?
How much evaluation history is reported?
Сведения для разработчиков
Автор
7alexhale5-rgbЛицензия
MIT
Ревизия Skillstore
r1
Примечание о версии
Автор не указал версию.
Ссылка
cabd63e41a4ad4b66103abc7e50c525e4710fb64
Актуальность поддержки
21.08.2026
Использование
1 загрузок · 0 просмотров
Структура файлов
📄 SKILL.md