model-per-chat
Assign Models Across Chats and Scheduled Tasks
Using one expensive model across every chat can exhaust limits, while careless downgrades can reduce quality. This skill guides model assignments using task importance, frequency, current plan rules, and quality checks.
Установить с помощью моего Агента
Скопируйте этот запрос в своего Агента. Он содержит каноническую страницу Skill и манифест.
Review the Skillstore skill "model-per-chat" from https://skillstore.io/skills/rabbitdesigns22-model-per-chat.md and its manifest at https://skillstore.io/api/skills/rabbitdesigns22-model-per-chat/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 «model-per-chat». My monitoring job runs every 20 minutes, while a supervisor approves customer messages. Where should I reduce costs?
Ожидаемый результат:
- Monitoring: evaluate a small model or script for routine checks, then verify quality and the next scheduled run.
- Supervisor: retain a stronger base model because mistakes affect customers.
- Status: recommendations only; no settings changes or measured savings reported.
Использование «model-per-chat». Prepare a cheaper setup for coding chats without weakening release checks.
Ожидаемый результат:
- Routine coding: consider a midrange base model with lower effort after representative checks pass.
- Hard debugging: delegate the difficult step after two failed attempts, then return to the base model.
- Release actions: require stronger-model review and existing owner approvals.
- Rollback: restore the stronger model if the reviewer returns two pieces of work within one week.
Аудит безопасности
Средний рискAll ten static findings are false positives: Markdown spans and model-selection guidance do not demonstrate shell execution or system reconnaissance. No evidence found of credential theft, exfiltration, or prompt injection in the reviewed file. One semantic concern remains: replaying recent tasks lacks explicit safeguards against repeating live side effects.
Подтверждённые проблемы безопасности (1)
Факторы риска
⚙️ Внешние команды (4)
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/rabbitdesigns22-model-per-chat/audits/1?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportЗначок Markdown
[](https://skillstore.io/skills/rabbitdesigns22-model-per-chat?utm_source=security_passport_badge)Значок HTML
<a href="https://skillstore.io/skills/rabbitdesigns22-model-per-chat?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/rabbitdesigns22-model-per-chat/security.svg" alt="Skillstore security assessment" loading="lazy"></a>Встраиваемая карточка
<iframe src="https://skillstore.io/embed/skills/rabbitdesigns22-model-per-chat.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
rabbitdesigns22. (2026). model-per-chat security audit report (audit version 1) [Author version unspecified]. Skillstore. https://skillstore.io/skills/rabbitdesigns22-model-per-chat/audits/1Цитата BibTeX
@techreport{rabbitdesigns22-rabbitdesigns22-model-per-chat-2026,
author = {rabbitdesigns22},
title = {model-per-chat security audit report (audit version 1)},
institution = {Skillstore},
year = {2026},
number = {1},
url = {https://skillstore.io/skills/rabbitdesigns22-model-per-chat/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: "model-per-chat security audit report (audit version 1)"
version: "unspecified"
type: report
authors:
- name: "rabbitdesigns22"
date-released: "2026-11-11"
url: "https://skillstore.io/skills/rabbitdesigns22-model-per-chat/audits/1"
identifiers:
- type: other
value: "skillstore:rabbitdesigns22-model-per-chat:audit:1"
description: "Skillstore immutable audit report identifier"
Оценка Skillstore
Почему такая оценка Достоверность доказательств: СреднийЧто вы можете построить
Balance Development Sessions
Separate routine coding work from complex debugging and release decisions, with stronger-model review where errors matter most.
Reduce Scheduled Monitoring Costs
Identify frequent jobs that can use smaller models or scripts, then verify the actual model used on their next run.
Protect Client Work Quality
Keep approval work on stronger models while assigning routine preparation to cheaper models with clear escalation and rollback rules.
Попробуйте эти промпты
Help inventory my chats and scheduled jobs. Ask for each purpose, current model, effort, run frequency, affected systems, and reviewer.
Review my chat inventory and current plan meters. Check official plan rules, score mistake consequences, and recommend models without changing settings.
Draft per-chat escalation rules for hard steps, repeated failures, and irreversible actions. Preserve owner approval requirements and mark unsupported delegation features.
Plan a batch downgrade with isolated replay checks, rollback criteria, and scheduled-run verification. Request approval before settings changes or live actions.
Лучшие практики
- Collect current usage meters and official plan rules before recommending assignments.
- Use isolated, read-only replay checks; obtain explicit approval before settings changes or repeated live actions.
- Verify actual session models after changes and document clear rollback criteria.
Избегать
- Downgrading every chat solely because a smaller model is cheaper.
- Replaying customer messages, spending, deployments, or deletion against live systems as a quality test.
- Claiming savings or completed switches without measurements or session verification.
Часто задаваемые вопросы
Does this skill switch every chat automatically?
Can it calculate exact subscription costs per turn?
How does it choose a model?
Does a scheduled task have its own model setting?
How should I test a cheaper model?
Is this an API model router for Claude or Codex?
Сведения для разработчиков
Автор
rabbitdesigns22Лицензия
MIT
Ревизия Skillstore
r1
Примечание о версии
Автор не указал версию.
Ссылка
f38ac41a6255e57978dd6f276aec539d5e9b5255
Актуальность поддержки
12.11.2026
Использование
0 загрузок · 0 просмотров
Структура файлов
📄 SKILL.md