К основному содержанию
правило CX Codex

disclaimer-language — правило для Claude Code

Правило disclaimer-language добавляет предупреждение к планам и результатам проверок, созданным ИИ-агентом. Оно напоминает, что выводы должны пройти проверку специалиста.

MIT ★ 1 627 обновлён 2026-10-25

Установить в GitHub Copilot · Вручную: положить файлы Все способы и платформы ↓
.github/instructions/disclaimer-language.instructions.md
Открыть на GitHub

Что делает disclaimer-language

Правило содержит тексты предупреждений для разных типов планирования и ревью: по безопасности, приватности, Responsible AI, цепочке поставок и другим направлениям. Агент показывает соответствующий текст при запуске или перед выводами. Предупреждение объясняет, что результат — вспомогательный материал, а не профессиональное заключение или одобрение.

В Claude Code файл правила размещают по пути .claude/rules/disclaimer-language.md. Область действия охватывает, среди прочего, планы и ревью в .copilot-tracking/ и документы в docs/planning/. Для юридических, security- и privacy-выводов правило прямо требует независимой проверки профильным специалистом: текст агента сам по себе не служит одобрением или сертификацией.

Какие задачи решает

  • добавить предупреждение к плану по безопасности
  • напомнить о проверке privacy-выводов специалистом
  • отделить AI-рекомендации по Responsible AI от юридического одобрения
  • предупредить, что оценки цепочки поставок требуют валидации

Примеры запросов агенту

  • Добавь предупреждение из правила к плану по безопасности.
  • Покажи оговорку перед выводами ревью приватности.
  • Укажи в плане Responsible AI, что рекомендации требуют независимой проверки.

Как установить disclaimer-language

Вручную: положить файлы

.github/instructions/disclaimer-language.instructions.md

Поле applyTo задаёт, к каким файлам применять

По шагам, с командами для каждой платформы: AGENTS.md и CLAUDE.md: что это, кто читает и какой выбрать

Вопросы о disclaimer-language

Как установить disclaimer-language в Claude Code?

Скопируйте файл правила в .claude/rules/disclaimer-language.md.

Что делает disclaimer-language?

Правило задаёт предупреждения для планов и ревью, подготовленных ИИ-агентами. Они поясняют, что выводы требуют проверки людьми с соответствующей квалификацией.

Заменяет ли правило проверку специалиста?

Нет. Для security- и privacy-выводов, а также оценок Responsible AI требуется независимая проверка; текст агента не считается одобрением или сертификацией.

Исходный текст

disclaimer-language.instructions.md · MIT
---
description: "Centralized disclaimer language for AI-assisted planning and review agents requiring professional review acknowledgment"
applyTo: '**/.copilot-tracking/rai-plans/**, **/.copilot-tracking/rai-reviews/**, **/.copilot-tracking/security-plans/**, **/.copilot-tracking/sssc-plans/**, **/.copilot-tracking/sssc-reviews/**, **/.copilot-tracking/performance-plans/**, **/.copilot-tracking/adr-plans/**, **/.copilot-tracking/dt/**, **/.copilot-tracking/ds/**, **/docs/planning/adrs/**, **/docs/planning/outcome-hypotheses/**, **/.copilot-tracking/reviews/code-reviews/**, **/.copilot-tracking/security/**, **/.copilot-tracking/accessibility/**, **/.copilot-tracking/privacy-plans/**, **/.copilot-tracking/privacy-reviews/**, **/.copilot-tracking/prd-sessions/**, **/.copilot-tracking/brd-sessions/**, **/.copilot-tracking/documentation/**'
---

# Disclaimer Language

Planning and review agents display a CAUTION disclaimer at startup or when presenting findings. Each H2 section below is the verbatim disclaimer for one planner or review family and is loaded by its consuming artifact before use.

<!--
Authoring contract (parsed by scripts/linting/Validate-PlannerArtifacts.ps1):
- Each planner or review family gets exactly one H2 section. The first whitespace-delimited word of the heading, lowercased, is the slug (e.g. "SSSC Planning" -> "sssc"; "Code-Review" -> "code-review").
- A section may override the derived slug with exactly one `<!-- disclaimer-id: lower-kebab -->` marker immediately before its CAUTION block. The identifier must use lowercase letters or digits separated by single hyphens. Explicit metadata takes precedence; first-word derivation applies only when metadata is absent. Empty, malformed, misplaced, or repeated markers are invalid.
- The slug derives: planner key `{slug}-planner`; disclaimer id `{slug}-full-disclaimer`; disclaimer label `{heading} Disclaimer`.
- Each H2 must contain exactly one `> [!CAUTION]` blockquote; only the first is extracted.
- The CAUTION prose should begin with `**Disclaimer:**` (trailing colon optional); the prefix is stripped before matching against artifact footers.
- Multi-line blockquote prose is joined with single spaces.
-->


## RAI Planning

> [!CAUTION]
> **Disclaimer:** This agent is an assistive tool only. It does not provide legal, regulatory, or compliance advice and does not replace Responsible AI review boards, ethics committees, legal counsel, compliance teams, or other qualified human reviewers. The output consists of suggested actions and considerations to support a user's own internal review and decision‑making. All RAI assessments, risk classification screenings, security models, and mitigation recommendations generated by this tool must be independently reviewed and validated by appropriate legal and compliance reviewers before use. Outputs from this tool do not constitute legal approval, compliance certification, or regulatory sign‑off.

## Security Planning

> [!CAUTION]
> **Disclaimer:** This agent is an assistive tool only. It does not provide legal, regulatory, or compliance advice and does not replace professional security review boards, penetration testing teams, compliance auditors, legal counsel, or other qualified human reviewers. The output consists of suggested actions and considerations to support a user's own internal security review and decision‑making. All security plans, threat models, security models, and mitigation recommendations generated by this tool must be independently reviewed and validated by appropriate security and compliance reviewers before use. Outputs from this tool do not constitute security approval, compliance certification, or regulatory sign‑off.

## Privacy Planning

> [!CAUTION]
> **Disclaimer:** This agent is an assistive tool only. It does not provide legal, regulatory, or privacy compliance advice and does not replace privacy counsel, data protection officers, compliance teams, or other qualified human reviewers. The output consists of suggested analyses, control recommendations, and privacy considerations to support a user's own internal privacy review and decision‑making. All privacy plans, DPIA assessments, data-flow analyses, and mitigation recommendations generated by this tool must be independently reviewed and validated by appropriate privacy and compliance reviewers before use. Outputs from this tool do not constitute legal approval, privacy certification, or regulatory sign‑off.

## Privacy Review

> [!CAUTION]
> **Disclaimer:** This agent is an assistive review tool only. It does not provide legal, regulatory, or privacy compliance sign-off and does not replace privacy counsel, data protection officers, compliance teams, or other qualified human reviewers. The output consists of AI-assisted findings, observations, and suggested next steps to support a reviewer's own privacy analysis and decision‑making. All privacy-review findings, DPIA observations, control recommendations, and risk assessments generated by this tool must be independently reviewed and validated by a qualified privacy reviewer before acting on them or treating any finding as resolved. Outputs from this tool do not constitute legal approval, privacy certification, or regulatory sign‑off.

## SSSC Planning

> [!CAUTION]
> **Disclaimer:** This agent is an assistive tool only. It does not provide legal, regulatory, or compliance advice and does not replace professional supply chain security review boards, OpenSSF Scorecard evaluators, SLSA auditors, legal counsel, or other qualified human reviewers. The output consists of suggested actions, review findings, and considerations to support a user's own internal supply chain security review and decision‑making. All supply chain assessments, review reports, gap analyses, backlog items, and mitigation recommendations generated by this tool must be independently reviewed and validated by appropriate security and compliance reviewers before use. Outputs from this tool do not constitute security approval, compliance certification, or regulatory sign‑off.

## Outcome-Hypothesis

> [!CAUTION]
> **Disclaimer:** This skill is an assistive decision-support tool only. It does not provide financial or professional investment advice and does not replace affected stakeholders, measurement owners, accountable sponsors or decision owners, or other qualified human reviewers. The investability verdict is an evidence-readiness signal only: "investable" means the defined evidence gates passed, and "not investable" means required evidence is incomplete. All scorecards, outcome hypotheses, investability verdicts, targets, and measurement plans must be independently reviewed and validated by affected stakeholders, the measurement owner, and the accountable sponsor or decision owner before funding, commitment, or implementation. Outputs from this tool do not constitute investment approval, funding authorization, stakeholder commitment, or measurement sign-off.

## ADR Planning

> [!CAUTION]
> **Disclaimer:** This agent is an assistive tool only. It does not provide legal, regulatory, architectural, or compliance advice and does not replace architecture review boards, design authorities, technical leadership, legal counsel, or other qualified human reviewers. The output consists of suggested decisions, considered options, consequences, and lineage metadata to support a user's own architecture decision-making. All Architecture Decision Records, supersession lineage, ASR trigger evaluations, and handoff work items generated by this tool must be independently reviewed and validated by appropriate architecture and engineering reviewers before adoption. Outputs from this tool do not constitute architectural approval, design sign-off, or compliance certification.

## PRD Requirements Planning

> [!CAUTION]
> **Disclaimer:** This agent is an assistive tool only. It does not provide product management approval, technical feasibility validation, or business sign-off and does not replace product managers, engineering leads, business stakeholders, or other qualified human reviewers. The output consists of suggested requirements, acceptance criteria, and product specifications to support a user's own product planning and decision-making. All Product Requirements Documents, functional requirements, non-functional requirements, and constraint definitions generated by this tool must be independently reviewed and validated by appropriate product and engineering reviewers before adoption. Outputs from this tool do not constitute product approval, requirements sign-off, or engineering commitment.

## BRD Requirements Planning

> [!CAUTION]
> **Disclaimer:** This agent is an assistive tool only. It does not provide business approval, regulatory compliance validation, or executive sign-off and does not replace business analysts, stakeholder representatives, compliance teams, or other qualified human reviewers. The output consists of suggested business requirements, objectives, and scope definitions to support a user's own business analysis and decision-making. All Business Requirements Documents, business objectives, stakeholder analysis, and requirement traceability generated by this tool must be independently reviewed and validated by appropriate business and compliance reviewers before adoption. Outputs from this tool do not constitute business approval, requirements sign-off, or stakeholder commitment.

## Design Thinking Coaching

> [!CAUTION]
> **Disclaimer:** This agent is an assistive coaching tool only. It does not conduct user research, observe stakeholders, or speak for the people whose problems you are designing for, and it does not replace primary research, direct stakeholder contact, design review, or product and strategy decision authority. Personas, problem statements, journey maps, empathy maps, concept tests, and other Design Thinking artifacts produced with this tool are scaffolding for your own research and synthesis — not substitutes for real stakeholder voice or observed behavior. Validate all AI-generated assumptions, personas, themes, and insights against actual stakeholders before treating any Design Thinking artifact as a basis for product, design, or strategy commitments. Outputs from this tool do not constitute validated research findings or design approval.

## Data Science and Engineering Coaching

<!-- disclaimer-id: data-science-engineering -->
> [!CAUTION]
> **Disclaimer:** This agent is an assistive data-science and data-engineering coaching tool only. It does not validate customer data, execute production pipelines, establish model fitness, or replace data owners, privacy and Responsible AI reviewers, engineering review, or business decision authority. Catalogs, feasibility findings, analyses, experiments, tests, and operational recommendations generated with this tool may be incomplete or inaccurate and must be independently reviewed against approved data sources, stakeholder evidence, and organizational controls before use. Outputs from this tool do not constitute data approval, feasibility sign-off, model approval, privacy or Responsible AI approval, or production readiness.

## Code-Review

> [!CAUTION]
> **Disclaimer:** This agent is an assistive review tool only. It does not provide engineering sign-off, security certification, or compliance approval and does not replace qualified human code review, security review boards, or other professional reviewers. The output consists of AI-assisted findings, observations, and suggested remediations to support a reviewer's own analysis and decision-making. All code-review findings — including functional, standards, and accessibility observations — generated by this tool must be independently reviewed and validated by a qualified human reviewer before acting on them, merging changes, or treating any finding as resolved. Outputs from this tool do not constitute engineering approval, merge authorization, or compliance certification.

## Security-Review

> [!CAUTION]
> **Disclaimer:** This agent is an assistive review tool only. It does not provide legal, regulatory, or compliance advice and does not replace professional security review boards, penetration testing teams, compliance auditors, or other qualified human reviewers. The output consists of AI-assisted findings, risk observations, and suggested remediations to support a reviewer's own security analysis and decision-making. All security-review findings, vulnerability assessments, and remediation recommendations generated by this tool must be independently reviewed and validated by a qualified security reviewer before acting on them or treating any finding as resolved. Outputs from this tool do not constitute security approval, compliance certification, or regulatory sign-off.

## Accessibility-Review

> [!CAUTION]
> **Disclaimer:** This agent is an assistive review tool only. It does not provide accessibility conformance certification or legal compliance sign-off and does not replace professional accessibility audits, assistive-technology user testing, or other qualified human reviewers. The output consists of AI-assisted findings, success-criteria observations, and suggested remediations to support a reviewer's own accessibility analysis and decision-making. All accessibility-review findings, conformance assessments, and remediation recommendations generated by this tool must be independently reviewed and validated by a qualified accessibility reviewer before acting on them or treating any finding as resolved. Outputs from this tool do not constitute accessibility conformance certification or legal compliance sign-off.

## Canonical session-state fields consumed by this disclaimer protocol

Planning and coaching agents that adopt this disclaimer protocol persist acknowledgment in their session state file (`state.json` for planners; `coaching-state.md` YAML for the DT Coach). The following field is required:

- `disclaimerShownAt` — ISO 8601 timestamp recording the most recent time the full disclaimer was shown to the user. Set to `null` until first display, leave unchanged during normal continuation, and update only when the full disclaimer is actually shown again. This field is not the immutable first-display audit record; the `noticeLog` array preserves the display history.

© microsoft · MIT · microsoft/hve-core

Похожие правила IDE

Все правила IDE по теме «Настройка агента: память, контекст, субагенты» →
ai-review Seth Hobson Команда ai-review для Claude Code проводит ревью кода с помощью статического анализа и ИИ. Она разбирает изменения и ищет уязвимости, проблемы производительности и архитектурные риски. CX Codex +1 ★ 42 780звёзд репо gdpr-data-mapping securityskills Скилл gdpr-data-mapping составляет реестр обработки персональных данных по GDPR, описывает маршруты передачи и регламенты инцидентов. CX Codex +1 239установок investigate-without-getting-made UseOSINT Скилл investigate-without-getting-made помогает не выдать себя при сборе информации: считает поверхность атрибуции, разводит рабочую и настоящую личность, подсказывает, когда хватит обычного браузера. CX Codex +1 58 200установок expression-language-injection yaklang Скилл expression-language-injection для Claude Code помогает находить и проверять уязвимости SpEL, OGNL и Java EL при аудите безопасности Java-веб-приложений. CX Codex +1 4 000установок Large Language Models Security Specialist majevva Промпт Large Language Models Security Specialist, который превращает чат в разбор безопасности LLM: ищет слабые места в ответах на тестовые запросы и предлагает, как закрыть риски. GP ChatGPT ★ 178 201звёзд репо data-privacy rohitg00 Плагин data-privacy для Claude Code, который находит персональные данные в коде и файлах, затем анонимизирует их. Запускается вручную через слэш-команду или автоматически перед коммитом. ★ 2 858звёзд репо

Из того же набора148

Весь набор →
markdown Задаёт стандарты разметки Markdown и схемы YAML frontmatter для всех файлов .md в проекте. Помогает агенту форматировать тексты без ошибок линтера. CX Codex +1 ★ 1 627звёзд репо System Architecture Reviewer Субагент System Architecture Reviewer разбирает архитектурные решения с учётом ограничений проекта, фиксирует компромиссы и помогает сохранить обоснование решений. Пригодится перед выбором архитектуры или оформлением ADR. CX Codex +1 ★ 1 627звёзд репо Data Science and Engineering Coach Субагент Data Science and Engineering Coach, который ведёт сессию data science и инженерии данных с явным выбором задач, сохранением состояния и маршрутизацией работы в специализированные скиллы. CX Codex +1 ★ 1 627звёзд репо RAI Planner Субагент RAI Planner проводит оценку этичности и надёжности ИИ-системы по стандарту NIST AI RMF 1.0, формируя итоговый отчёт rai-plan.md через шесть фаз с проверками. CX Codex +1 ★ 1 627звёзд репо vally-tests Скилл vally-tests генерирует тесты Vally, проверяющие, что промпты, инструкции, субагенты и скиллы ведут себя так, как заявлено. Пригодится при приёмке новых артефактов в репозиторий. CX Codex +1 ★ 1 627звёзд репо functional-planner Скилл functional-planner разбирает документ требований на структуру задач для Azure DevOps, GitHub или Jira без прямых правок в трекере. CX Codex +1 ★ 1 627звёзд репо backlog-execute Скилл backlog-execute для Claude Code создаёт задачи или применяет handoff в Azure DevOps, GitHub и Jira. Вносит согласованные изменения в трекер по запросу или по файлу операций. CX Codex +1 ★ 1 627звёзд репо backlog-templates Скилл backlog-templates для Claude Code хранит общие шаблоны задач для ADO и GitHub с метаданными, ID и правилами санитизации. Их используют планировщики RAI, Security, SSSC, Accessibility и Privacy. CX Codex +1 ★ 1 627звёзд репо

Спонсоры проекта

Компании, которые помогают держать аналитику открытой