📦

huashu-design

Name: huashu-design
Author: alchaincyf

v2.0 Ревизия содержимого r1 Высокий риск 🌐 Доступ к сети🔑 Переменные окружения📁 Доступ к файловой системе⚙️ Внешние команды⚡ Содержит скрипты

Create HTML Prototypes, Slides, and Animations

Visual briefs need clear direction, consistent assets, and practical export options. Huashu Design guides three visual drafts, HTML implementation, local verification, and optional presentation or video exports.

Поддерживает: Claude Codex Code(CC)

⚠️ 38 Плохо

Maintain this Skill? Claim it ⚠️ 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 "huashu-design" from https://skillstore.io/skills/alchaincyf-huashu-design.md and its manifest at https://skillstore.io/api/skills/alchaincyf-huashu-design/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/alchaincyf-huashu-design.md Подписанный manifest GET /api/skills/alchaincyf-huashu-design/manifest Подписанный lockfile GET /api/skills/alchaincyf-huashu-design/lockfile

Протестировать

Использование «huashu-design». Create an onboarding prototype for a reading application using supplied screenshots.

Ожидаемый результат:

Three visual direction screenshots for selection.
An approved HTML prototype with interactive onboarding screens.
A verification summary covering key clicks and remaining limitations.

Использование «huashu-design». Turn an approved workshop outline into an eight-slide presentation.

Ожидаемый результат:

Representative slides in three distinct visual directions.
A selected HTML deck with overview navigation and a PDF handout.
An editable PPTX when requested, with export fidelity notes.

Использование «huashu-design». Produce a short product demonstration from approved interface images.

Ожидаемый результат:

Three direction boards followed by an approved storyboard.
An HTML animation and MP4 with planned music and sound effects.
A quality checklist covering framing, text, timing, and audio.

Аудит безопасности

Высокий риск

v1 • 13.10.2026 Открыть версионный отчет

All 400 supplied findings were adjudicated; most match benign design text, dependency checksums, or declared local tooling. Retained risks concern Python execution, cloud consent and transport, proxy bypass, browser credentials, unescaped labels, destructive cleanup, and global installation. No evidence found of deliberate malware or auditor-targeted prompt injection; 1,337 omitted matches still require manual review before automatic publication. Static review was capped at 400/1737 representative findings; omitted static matches are unconfirmed, so automatic publishing stays disabled until manual review.

125

Просканировано файлов

52,622

Проанализировано строк

Пункты проверки

Ложные срабатывания проигнорированы

Подтверждённые проблемы безопасности (4)

Высокий

scripts/sfx-cues.sh:24-29 scripts/verify-video.sh:56-65

Media Parameters Become Executable Python

Cue timestamps and verification options are interpolated into Python source without validation. Crafted cue files or parameter values can execute code with user privileges.

Source inspection directly shows unvalidated TSV text and command-line values inserted into Python expressions or quoted literals. No exploit execution was required or performed.

Средний

scripts/narrate-pipeline.mjs:203-207 scripts/narrate-pipeline.mjs:347-348 SECURITY.md:36

Cleanup Can Delete Existing User Files

The narration pipeline reuses an existing .tmp directory beneath the requested output directory, then recursively deletes it. Existing unrelated files can be lost.

Recursive creation accepts existing directories, and cleanup deletes the whole directory without ownership checks. This contradicts the claimed unique, script-owned scratch guarantee.

Средний

scripts/cloud/ai-review-video.py:269-286 scripts/cloud/ai-review-video.py:199-205 scripts/cloud/ai-review-video.py:152-160 SECURITY.md:11

Cloud Consent Omits Director-Note Uploads

The consent message describes compressed video uploads, but optional director notes also enter remote prompts. Confidential planning text can leave without specific disclosure.

The context file is read into prompts submitted through session.post. The consent notice and destination table describe video, but do not enumerate director-note text.

Средний

SKILL.md:392 references/hyperframes-backend.md:22-34

Default Workflow Trusts an Unpinned Initializer

The recommended npx initializer uses automatic acceptance without a version pin. Third-party executable code can change before project-level pinning and install additional agent instructions.

The command uses npx -y without a version; only the generated project receives a version pin. Upstream package behavior was not executed or verified.

Пункты проверки возможностей (20)

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

Высокий

scripts/verify-video.sh:48

Dynamic code evaluation with eval()

FPS=$(python3 -c "print(round(eval('${FPS_RAW:-0}' if '${FPS_RAW:-0}'!='0/0' else '0'),2))" 2>/dev/n

Python eval parses ffprobe output as executable source, but normal frame-rate output is a numeric fraction. No direct attacker-controlled path was established; retained conservatively.

Высокий

references/hyperframes-backend.md:23

Hidden file in home directory

> `~/.claude/skills/`**(渲染后端的合成契约文档,纯文档无可执行 hook)。介意的话先跑

The documented initializer installs 19 instruction skills into the global agent directory. These changes expand persistent instruction trust beyond the current project.

Высокий

references/hyperframes-backend.md:31

Hidden file in home directory

生成 index.html / hyperframes.json / meta.json / package.json(pin 了 CLI 版本)+ 项目级 CLAUDE.md。init 会把 19

The documented initializer installs 19 instruction skills into the global agent directory. These changes expand persistent instruction trust beyond the current project.

Высокий

scripts/cloud/tts-doubao.mjs:195

Environment variable access (dot notation)

const endpoint = process.env.DOUBAO_TTS_ENDPOINT || 'https://openspeech.bytedance.com/api/v3/tts/uni

Only the initial hostname is checked; HTTPS and redirect destinations are not enforced. Custom authentication headers can leak through plaintext transport or unsafe redirects.

Высокий

scripts/cloud/tts-doubao.mjs:221

Fetch API call

const res = await fetch(endpoint, {

Only the initial hostname is checked; HTTPS and redirect destinations are not enforced. Custom authentication headers can leak through plaintext transport or unsafe redirects.

Высокий

scripts/cloud/tts-doubao.mjs:195

Hardcoded URL

const endpoint = process.env.DOUBAO_TTS_ENDPOINT || 'https://openspeech.bytedance.com/api/v3/tts/uni

Only the initial hostname is checked; HTTPS and redirect destinations are not enforced. Custom authentication headers can leak through plaintext transport or unsafe redirects.

Средний

references/hyperframes-backend.md:23

Hidden file access

> `~/.claude/skills/`**(渲染后端的合成契约文档,纯文档无可执行 hook)。介意的话先跑

The initializer is documented to write additional skills into a global instruction directory. This creates persistent trust changes outside the project.

Средний

references/hyperframes-backend.md:31

Hidden file access

生成 index.html / hyperframes.json / meta.json / package.json(pin 了 CLI 版本)+ 项目级 CLAUDE.md。init 会把 19

The initializer is documented to write additional skills into a global instruction directory. This creates persistent trust changes outside the project.

Средний

assets/deck_index.html:307

innerHTML assignment (XSS risk)

counter.innerHTML = (idx + 1) + ' / ' + deck.length + ' <span class="label">' + (deck[idx].label ||

Labels enter innerHTML without escaping, allowing scriptable markup if untrusted titles are reused. The bundled manifest is inline and benign; external attacker control is unproven.

Средний

scripts/cloud/tts-doubao.mjs:44

Node.js fs operations

const text = fs.readFileSync(envPath, 'utf8');

The loader imports arbitrary configuration keys into process.env before consent checks. A configuration file can set HUASHU_CLOUD_OK=1 and bypass invocation approval.

Средний

references/ai-video-review.md:19

Proxy environment variable

cd 项目目录 && unset ALL_PROXY # 脚本内已免疫代理,unset是双保险

This code or instruction disables inherited proxy routing. Requests can bypass required enterprise inspection or privacy controls despite the stated TLS troubleshooting purpose.

Средний

references/brand-asset-protocol.md:77

Proxy environment variable

unset ALL_PROXY HTTP_PROXY HTTPS_PROXY all_proxy http_proxy https_proxy # 清代理,否则 TLS 易炸

This code or instruction disables inherited proxy routing. Requests can bypass required enterprise inspection or privacy controls despite the stated TLS troubleshooting purpose.

Средний

scripts/cloud/ai-review-video.py:31

Proxy environment variable

代理:requests session 关闭 trust_env(不继承本机代理配置),免疫 ALL_PROXY 之类残留代理导致的 TLS 报错。

This code or instruction disables inherited proxy routing. Requests can bypass required enterprise inspection or privacy controls despite the stated TLS troubleshooting purpose.

Средний

scripts/cloud/ai-review-video.py:286

Proxy environment variable

session.trust_env = False # 免疫 ALL_PROXY 等代理坑

This code or instruction disables inherited proxy routing. Requests can bypass required enterprise inspection or privacy controls despite the stated TLS troubleshooting purpose.

Средний

scripts/fetch_images.py:18

Proxy environment variable

for _k in ("ALL_PROXY", "all_proxy", "HTTP_PROXY", "http_proxy", "HTTPS_PROXY", "https_proxy"):

This code or instruction disables inherited proxy routing. Requests can bypass required enterprise inspection or privacy controls despite the stated TLS troubleshooting purpose.

Средний

scripts/cloud/tts-doubao.mjs:44

Synchronous file operations

const text = fs.readFileSync(envPath, 'utf8');

The loader imports arbitrary configuration keys into process.env before consent checks. A configuration file can set HUASHU_CLOUD_OK=1 and bypass invocation approval.

Средний

references/react-setup.md:154

Fetch API call

const res = await fetch('https://api.anthropic.com/v1/messages', {

The optional example exposes the entered API key to page scripts and the visible DOM. Local-demo warnings and recommended mocks reduce usage, not credential exposure.

Низкий

scripts/cloud/tts-doubao.mjs:55

Environment variable access (bracket notation)

if (!(key in process.env)) process.env[key] = val;

The loader imports arbitrary configuration keys into process.env before consent checks. A configuration file can set HUASHU_CLOUD_OK=1 and bypass invocation approval.

Низкий

scripts/cloud/tts-doubao.mjs:55

Environment variable object

if (!(key in process.env)) process.env[key] = val;

The loader imports arbitrary configuration keys into process.env before consent checks. A configuration file can set HUASHU_CLOUD_OK=1 and bypass invocation approval.

Низкий

scripts/fetch_images.py:19

Python environment access

os.environ.pop(_k, None)

This code or instruction disables inherited proxy routing. Requests can bypass required enterprise inspection or privacy controls despite the stated TLS troubleshooting purpose.

Факторы риска

🌐 Доступ к сети (50)

🔑 Переменные окружения (50)

📁 Доступ к файловой системе (50)

.gitignore:3 .gitignore:7 assets/narration_stage.jsx:352 assets/narration_stage.jsx:424 assets/narration_stage.jsx:440 assets/personal-asset-index.example.json:4 assets/personal-asset-index.example.json:4 demos/hero-animation-v10-en.html:970 demos/md-html-narration/md-html-demo.html:256 demos/w3-fallback-advisor-en.html:350 demos/w3-fallback-advisor-en.html:361 demos/w3-fallback-advisor-en.html:372 demos/w3-fallback-advisor.html:357 demos/w3-fallback-advisor.html:369 demos/w3-fallback-advisor.html:381 README.en.md:58 README.en.md:67 README.en.md:58 README.en.md:67 README.md:70 README.md:79 README.md:70 README.md:79 references/ai-video-review.md:20 references/ai-video-review.md:20 references/animation-pitfalls.md:365 references/brand-asset-protocol.md:79 references/camera-language.md:198 references/cinematic-patterns.md:179 references/editable-pptx.md:243 references/hero-animation-case-study.md:5 references/hyperframes-backend.md:23 references/hyperframes-backend.md:31 references/hyperframes-backend.md:42 references/hyperframes-backend.md:43 references/hyperframes-backend.md:23 references/hyperframes-backend.md:31 references/hyperframes-backend.md:42 references/hyperframes-backend.md:43 references/sfx-library.md:226 references/slide-decks.md:225 references/slide-decks.md:257 references/slide-decks.md:321 references/slide-decks.md:175 references/slide-decks.md:175 references/verification.md:58 references/verification.md:58 scripts/add-music.sh:7 scripts/add-music.sh:36 scripts/cloud/ai-review-video.py:40

⚙️ Внешние команды (50)

⚡ Содержит скрипты (23)

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/alchaincyf-huashu-design/audits/1?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Значок Markdown

[![Skillstore security assessment](https://skillstore.io/badges/skills/alchaincyf-huashu-design/security.svg)](https://skillstore.io/skills/alchaincyf-huashu-design?utm_source=security_passport_badge)

Значок HTML

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

Встраиваемая карточка

<iframe src="https://skillstore.io/embed/skills/alchaincyf-huashu-design.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

alchaincyf. (2026). huashu-design security audit report (audit version 1) [Author version 2.0]. Skillstore. https://skillstore.io/skills/alchaincyf-huashu-design/audits/1

Цитата BibTeX

@techreport{alchaincyf-alchaincyf-huashu-design-2026,
  author = {alchaincyf},
  title = {huashu-design security audit report (audit version 1)},
  institution = {Skillstore},
  year = {2026},
  number = {1},
  url = {https://skillstore.io/skills/alchaincyf-huashu-design/audits/1},
  note = {Author version 2.0}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "huashu-design security audit report (audit version 1)"
version: "2.0"
type: report
authors:
  - name: "alchaincyf"
date-released: "2026-10-13"
url: "https://skillstore.io/skills/alchaincyf-huashu-design/audits/1"
identifiers:
  - type: other
    value: "skillstore:alchaincyf-huashu-design:audit:1"
    description: "Skillstore immutable audit report identifier"

Оценка Skillstore

Почему такая оценка Достоверность доказательств: Средний

Архитектура

Сопровождаемость

Контент

Сообщество

Соответствие спецификации

Что вы можете построить

Compare Product Interface Directions

Explore three branded interface drafts, select a direction, and build an interactive prototype for stakeholder review.

Build Training Presentations

Turn approved teaching material into navigable HTML slides, a PDF handout, and an optional editable PowerPoint deck.

Produce Product Demonstration Videos

Plan keyframes, animate approved product assets, and export videos with narration, captions, music, and sound effects.

Попробуйте эти промпты

Explore Three Design Directions

Create three HTML infographic drafts about [topic] for [audience]. Use [approved content]. Show screenshots and wait for my selection.

Build an Interactive Prototype

Prototype [product] using [brand assets] and [key flows]. Show three directions first. After approval, implement interactions and verify clicks locally.

Create a Presentation Package

Design a [slide count]-slide deck from [outline] for [audience]. Show three representative directions. After approval, deliver HTML and PDF. Ask before PPTX export.

Direct a Product Animation

Plan a [duration]-second animation using [approved assets]. Show three direction boards, then storyboard the selected direction. Export MP4 with captions and audio. Keep cloud tools disabled and request approval before external installation.

Лучшие практики

Provide approved content, brand assets, audience details, and output dimensions before selecting a direction.
Verify screenshots, interactions, typography, and exported files before delivery.
Fix reported issues before using bundled scripts in an isolated project. Require explicit approval for cloud transfers and global skill installation.

Избегать

Use the skill as a substitute for production application engineering or backend architecture.
Run bundled scripts on untrusted cue tables or parameter values before fixing the reported injection risks.
Upload confidential notes or enter production API keys into browser demos without reviewing data exposure.

Часто задаваемые вопросы

Which assistants can use this skill?

The report lists Claude, Codex, and Claude Code. Tool access and available runtime features determine which workflows can run.

Does every task require three design directions?

New visual designs normally start with three drafts. Explicit user waivers, approved-direction iterations, and mechanical export tasks are documented exceptions.

Can it export editable PowerPoint presentations?

Yes. Supplied scripts support constrained HTML conversion and rendered-layout conversion. Review editability and visual fidelity after export.

Do I need API keys?

Local design and export scripts do not require API keys. Optional cloud narration and video review require your credentials and approval.

Will every workflow run offline?

No. Asset searches, font downloads, package installation, update checks, and optional cloud tools can use the network.

What security precautions are required?

Fix reported script risks and use trusted inputs with isolated output directories. Review uploaded text, videos, and global skill installations before approval.