Навыки extension-inference

📦

extension-inference

Name: extension-inference
Author: caffeinelabs

v0.1.0 Ревизия содержимого r1 Высокий риск ⚙️ Внешние команды🌐 Доступ к сети📁 Доступ к файловой системе

Build Secure Caffeine Inference Features

Caffeine applications need LLM access without exposing provider keys or misconfiguring canister outcalls. This skill provides the supported authenticated Motoko integration pattern.

Поддерживает: Claude Codex Code(CC)

⚠️ 38 Плохо

Остановитесь и запросите подтверждение перед установкой.

Проверьте план и получите явное согласие пользователя перед изменением файлов.

Установить с помощью моего Агента

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

Запрос агента

Review the Skillstore skill "extension-inference" from https://skillstore.io/skills/caffeinelabs-extension-inference.md and its manifest at https://skillstore.io/api/skills/caffeinelabs-extension-inference/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.

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/caffeinelabs-extension-inference.md Подписанный manifest GET /api/skills/caffeinelabs-extension-inference/manifest Подписанный lockfile GET /api/skills/caffeinelabs-extension-inference/lockfile

Протестировать

Использование «extension-inference». Add a summarization feature to my Caffeine application without asking users for an API key.

Ожидаемый результат:

A backend integration plan using platform credentials, a router-based chat request, validated response handling, and a simple pending frontend state.

Использование «extension-inference». Check whether my canister inference method is ready for production.

Ожидаемый результат:

Credential handling is backend-only and refreshed for each request.
Authorization, quotas, rate limits, and prompt-size limits are required before deployment.
The request must remain non-replicated and cycle usage should be monitored.

Использование «extension-inference». Can this integration generate images and stream partial responses?

Ожидаемый результат:

No. This host supports chat completions with complete responses, not image generation or streaming.

Аудит безопасности

Высокий риск

v1 • 29.08.2026 Открыть версионный отчет

All 59 static findings are false positives caused by Markdown code formatting, documentation links, and relative links. However, the canonical public chat endpoint lacks authorization, quotas, and rate limits, creating a high-confidence resource-abuse risk.

Просканировано файлов

191

Проанализировано строк

Пункты проверки

Ложные срабатывания проигнорированы

Подтверждённые проблемы безопасности (1)

Высокий

SKILL.md:81-84 SKILL.md:72-73

Unauthenticated Paid Inference Endpoint

The canonical actor exposes chat as a public shared method without caller authorization, quotas, or rate limits. Any caller could consume inference capacity and cycles.

The example directly exposes a public paid outcall and the document identifies inference spend, but it provides no abuse controls or caller checks.

Факторы риска

⚙️ Внешние команды (50)

🌐 Доступ к сети (2)

SKILL.md:23 SKILL.md:138

📁 Доступ к файловой системе (2)

SKILL.md:36 SKILL.md:155

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/caffeinelabs-extension-inference/audits/1?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Значок Markdown

[![Skillstore security assessment](https://skillstore.io/badges/skills/caffeinelabs-extension-inference/security.svg)](https://skillstore.io/skills/caffeinelabs-extension-inference?utm_source=security_passport_badge)

Значок HTML

<a href="https://skillstore.io/skills/caffeinelabs-extension-inference?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/caffeinelabs-extension-inference/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Встраиваемая карточка

<iframe src="https://skillstore.io/embed/skills/caffeinelabs-extension-inference.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

caffeinelabs. (2026). extension-inference security audit report (audit version 1) [Author version 0.1.0]. Skillstore. https://skillstore.io/skills/caffeinelabs-extension-inference/audits/1

Цитата BibTeX

@techreport{caffeinelabs-caffeinelabs-extension-inference-2026,
  author = {caffeinelabs},
  title = {extension-inference security audit report (audit version 1)},
  institution = {Skillstore},
  year = {2026},
  number = {1},
  url = {https://skillstore.io/skills/caffeinelabs-extension-inference/audits/1},
  note = {Author version 0.1.0}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "extension-inference security audit report (audit version 1)"
version: "0.1.0"
type: report
authors:
  - name: "caffeinelabs"
date-released: "2026-08-29"
url: "https://skillstore.io/skills/caffeinelabs-extension-inference/audits/1"
identifiers:
  - type: other
    value: "skillstore:caffeinelabs-extension-inference:audit:1"
    description: "Skillstore immutable audit report identifier"

Оценка Skillstore

Почему такая оценка Достоверность доказательств: Средний

Архитектура

100

Сопровождаемость

Контент

Сообщество

100

Соответствие спецификации

Что вы можете построить

Add a Canister Chat Endpoint

Create a Motoko backend method that sends prompts through Caffeine Inference and returns validated text.

Review Inference Architecture

Check that credentials remain backend-only and requests use supported configuration, model, and replication settings.

Plan an AI Product Feature

Confirm whether chat completions meet product requirements before designing frontend states and backend interfaces.

Попробуйте эти промпты

Add Basic Chat

Add a basic Caffeine Inference chat method to my Motoko canister. Use platform credentials and explain each changed file.

Integrate Existing Backend

Integrate Caffeine Inference into this existing Motoko backend. Preserve current architecture and handle empty or missing response content.

Review Security and Cost Controls

Review this Caffeine Inference endpoint for credential exposure, replicated outcalls, unauthorized usage, cycle costs, and missing request limits.

Design Production Inference Workflow

Design a production Caffeine Inference workflow with authorization, per-principal quotas, prompt limits, usage accounting, retries, and user-facing failure states.

Лучшие практики

Create configuration from the platform environment for every request and never persist its authorization data.
Keep non-replicated outcalls enabled and use the router model for every chat completion.
Protect paid inference methods with authorization, quotas, rate limits, and bounded prompt sizes.

Избегать

Do not collect OpenAI keys or expose inference credentials through frontend settings, logs, queries, or actor state.
Do not call provider endpoints directly or replace platform configuration with hand-built HTTP requests.
Do not expose an unrestricted public inference method that allows anonymous callers to consume cycles.

Часто задаваемые вопросы

Do users need to provide an OpenAI API key?

No. Caffeine provisions inference credentials and the canister reads them at runtime.

Which model should requests use?

Use the router model. The platform selects an appropriate underlying model for each request.

Can the frontend call the inference host directly?

No. The frontend should call a backend canister method so credentials never reach the browser.

Does this skill support streaming?

No. Management-canister HTTP returns the complete response body, so streaming must remain disabled.

Which inference APIs are available?

Chat completions and the model catalog are documented. Embeddings, images, audio, files, moderation, and other OpenAI APIs are unavailable.

What production controls should I add?

Add caller authorization, quotas, rate limits, prompt-size limits, usage accounting, and clear retry behavior.

Сведения для разработчиков

Автор

caffeinelabs

Лицензия

MIT

Версия автора

v0.1.0

Ревизия Skillstore

Репозиторий

https://github.com/caffeinelabs/skills/tree/38a113689f275bb2efe5fa87e4b151c4a8899e8d/skills/extension-inference

Ссылка

2197f31c95ebfaf5ee9b50f95c8ecb53afcd7380

Актуальность поддержки

29.08.2026

Использование

1 загрузок · 0 просмотров

Структура файлов

📄 SKILL.md