Навыки fact-check-x-complete

📦

fact-check-x-complete

Name: fact-check-x-complete
Author: asi2030

v1.1.5 Ревизия содержимого r1 Высокий риск 🌐 Доступ к сети🔑 Переменные окружения📁 Доступ к файловой системе⚙️ Внешние команды⚡ Содержит скрипты

Verify AI Answers Across Multiple Platforms

AI platforms can provide conflicting claims and uneven citations. This skill captures their answers, compares facts, and produces traceable verification reports.

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 "fact-check-x-complete" from https://skillstore.io/skills/asi2030-fact-check-x-complete.md and its manifest at https://skillstore.io/api/skills/asi2030-fact-check-x-complete/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/asi2030-fact-check-x-complete.md Подписанный manifest GET /api/skills/asi2030-fact-check-x-complete/manifest Подписанный lockfile GET /api/skills/asi2030-fact-check-x-complete/lockfile

Протестировать

Использование «fact-check-x-complete». Compare two platforms on a housing policy question and stop after claim comparison.

Ожидаемый результат:

A capture report preserves both answers and citations. A comparison report separates shared claims, conflicting figures, and missing evidence.

Использование «fact-check-x-complete». Complete an authoritative verification for several disputed claims.

Ожидаемый результат:

A verified answer includes only claims supported by authoritative evidence.
Unresolved claims appear in a separate evidence-gap section.
A final report compares platform coverage, citation quality, and verified outcomes.

Аудит безопасности

Высокий риск

v1 • 27.08.2026 Открыть версионный отчет

Most static matches are benign test fixtures, documentation, fixed-argument subprocesses, report templates, and standard integrity data. Confirmed issues include unrestricted trusted-search endpoint overrides that can redirect an API key, declared third-party query transmission, and indirect prompt-injection exposure when the host interprets captured content. Static review was capped at 400/947 representative findings; omitted static matches are unconfirmed, so automatic publishing stays disabled until manual review.

Просканировано файлов

15,362

Проанализировано строк

Пункты проверки

Ложные срабатывания проигнорированы

Подтверждённые проблемы безопасности (1)

Высокий

SKILL.md:174-178 SKILL.md:219-221

Indirect Prompt Injection Exposure in Collected Content

The workflow directs the host agent to interpret captured platform answers and external evidence, then write analysis files. It does not define those materials as untrusted data or instruct the agent to ignore embedded commands, so malicious content could influence analysis and subsequent tool use.

The cited workflow explicitly assigns semantic interpretation of captured answers and retrieved evidence to the host agent. The complete skill instructions contain no untrusted-content or embedded-instruction boundary.

Пункты проверки возможностей (9)

Это реальные локальные возможности, которые могут ожидаться для этого навыка, поэтому они требуют проверки, но не считаются подтверждённым вредоносным поведением.

Средний

modules/fact-check-x-authoritative-verify/scripts/authority_verify.py:110

getenv function call

endpoint = os.getenv("FACTCHECK_TRUSTED_SEARCH_URL", "https://open.dknowc.cn/dependable/search").str

The configured endpoint is used for requests carrying the trusted-search API key, but the value is not restricted to HTTPS or an approved host. A redirected endpoint could receive the credential and query.

Средний

scripts/trusted_search_config.py:115

getenv function call

endpoint = os.getenv("FACTCHECK_TRUSTED_SEARCH_URL", DEFAULT_ENDPOINT).strip()

The validation endpoint can be replaced without scheme or host restrictions before the API key is transmitted. A hostile configured endpoint could collect the credential.

Средний

modules/fact-check-x-authoritative-verify/scripts/authority_verify.py:110

Hardcoded URL

endpoint = os.getenv("FACTCHECK_TRUSTED_SEARCH_URL", "https://open.dknowc.cn/dependable/search").str

Although the default is HTTPS, the endpoint can be replaced without scheme or host validation before an API key is sent. This creates a credential-redirection risk.

Средний

modules/fact-check-x-authoritative-verify/scripts/authority_verify.py:110

Python getenv function

endpoint = os.getenv("FACTCHECK_TRUSTED_SEARCH_URL", "https://open.dknowc.cn/dependable/search").str

Средний

scripts/trusted_search_config.py:115

Python getenv function

endpoint = os.getenv("FACTCHECK_TRUSTED_SEARCH_URL", DEFAULT_ENDPOINT).strip()

The validation endpoint can be replaced without scheme or host restrictions before the API key is transmitted. A hostile configured endpoint could collect the credential.

Низкий

modules/fact-check-x-authoritative-verify/scripts/authority_verify.py:114

Python HTTP libraries

request = urllib.request.Request(endpoint, data=json.dumps(payload, ensure_ascii=False).encode("utf-

This request transmits an atomic verification query and API key to the configured trusted-search service. The transfer is declared functionality, but it is a real third-party data exposure.

Низкий

modules/fact-check-x-authoritative-verify/scripts/authority_verify.py:116

Python HTTP libraries

with urllib.request.urlopen(

The URL open call performs the declared outbound trusted-search request containing a query and credential header. This is intentional but remains a real network exposure.

Низкий

scripts/trusted_search_config.py:122

Python HTTP libraries

request = urllib.request.Request(

The request includes the trusted-search API key and a validation query. This is expected configuration behavior, but it exposes both values to the selected endpoint.

Низкий

scripts/trusted_search_config.py:129

Python HTTP libraries

with urllib.request.urlopen(

The URL open call sends the trusted-search credential to an external validation service. The action is intentional, but the network disclosure is real.

Факторы риска

🌐 Доступ к сети (50)

🔑 Переменные окружения (50)

📁 Доступ к файловой системе (50)

⚙️ Внешние команды (50)

⚡ Содержит скрипты (13)

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/asi2030-fact-check-x-complete/audits/1?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Значок Markdown

[![Skillstore security assessment](https://skillstore.io/badges/skills/asi2030-fact-check-x-complete/security.svg)](https://skillstore.io/skills/asi2030-fact-check-x-complete?utm_source=security_passport_badge)

Значок HTML

<a href="https://skillstore.io/skills/asi2030-fact-check-x-complete?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/asi2030-fact-check-x-complete/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Встраиваемая карточка

<iframe src="https://skillstore.io/embed/skills/asi2030-fact-check-x-complete.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

asi2030. (2026). fact-check-x-complete security audit report (audit version 1) [Author version 1.1.5]. Skillstore. https://skillstore.io/skills/asi2030-fact-check-x-complete/audits/1

Цитата BibTeX

@techreport{asi2030-asi2030-fact-check-x-complete-2026,
  author = {asi2030},
  title = {fact-check-x-complete security audit report (audit version 1)},
  institution = {Skillstore},
  year = {2026},
  number = {1},
  url = {https://skillstore.io/skills/asi2030-fact-check-x-complete/audits/1},
  note = {Author version 1.1.5}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "fact-check-x-complete security audit report (audit version 1)"
version: "1.1.5"
type: report
authors:
  - name: "asi2030"
date-released: "2026-08-27"
url: "https://skillstore.io/skills/asi2030-fact-check-x-complete/audits/1"
identifiers:
  - type: other
    value: "skillstore:asi2030-fact-check-x-complete:audit:1"
    description: "Skillstore immutable audit report identifier"

Оценка Skillstore

Почему такая оценка Достоверность доказательств: Средний

Архитектура

100

Сопровождаемость

Контент

Сообщество

Соответствие спецификации

Что вы можете построить

Compare Research Answers

Compare answers and citations from selected AI platforms before using them in a research brief.

Audit Decision Evidence

Trace disputed operational claims to captured sources and authoritative evidence before making a business decision.

Review AI Reliability

Evaluate answer coverage, citation faithfulness, and verified accuracy across supported AI services.

Попробуйте эти промпты

Check One Answer

Fact-check this question on DeepSeek: [question]. Capture the full answer and citations. Stop after the first report.

Compare Two Platforms

Compare DeepSeek and Doubao on this question: [question]. Show agreements, conflicts, citations, and evidence gaps.

Run Full Verification

Run the complete workflow for [question] across [platforms]. Continue through authoritative verification and deliver every staged report.

Audit a Complex Claim Set

Verify [question] across [platforms]. Split compound statements into atomic claims. Preserve conflicting evidence and exclude unresolved claims from final conclusions.

Лучшие практики

Choose only platforms relevant to the question and confirm each capture before continuing.
Review captured answers and citations before requesting authoritative verification.
Inspect reports for confidential information before sharing the final package.

Избегать

Do not treat the unverified comparison draft as an authoritative answer.
Do not bypass failed captures with partial results or substituted search summaries.
Do not paste credentials into prompts, reports, commands, or shared artifacts.