← Back to Skills Marketplace
bozoyan

design-ads 海报制作

by bozoyan · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
108
Downloads
1
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install design-ads
Description
AI 知识类海报/封面图设计生成器。根据用户提供的主题内容,自动生成 1200x1800 像素的竖版海报图片, 风格统一为深色科技感设计(暗黑背景 + 渐变强调色 + 思源宋体标题 + 思源黑体正文)。 支持 6 种模板布局:3 种封面样式(居中图标型、特性网格型、列表展示型)和 3 种内页样式(工具详情型、提示...
README (SKILL.md)

Design Ads - AI 知识海报设计器

根据 references/design-system.md 中的完整设计规范,为用户生成专业级 AI 知识海报。

核心原则

画布空间利用

┌─ 1200px 宽 ─────────────────────────────┐
│120px│    960px 内容区 (必须充分利用)     │120px
│     │                                    │
│     │   封面:大字 + 留白 = 视觉冲击     │
│     │   内页:内容充实 + 排版整齐         │
│     │                                    │
│150px│                                    │150px
└─────┴────────────────────────────────────┘──
        ↑
    1800px 高,内容区 1500px

铁律:内容必须在 960x1500px 内容区内合理分布,不能集中在顶部一小块区域。

  • 封面:用超大字号 + 合理留白撑满纵向空间,元素垂直均匀分布
  • 内页:内容要充实,从标签区到底部分隔线之间填满,不留大块空白

封面 vs 内页的区别

封面(3种) 内页(3种)
信息量 精简 — 1个主标题+1行描述+图标/网格 充实 — 标题+多段描述+列表/代码块
字号策略 超大标题占视觉主体(160-175px) 标题大但内容更多(95-110px)
目的 吸引点击、传递核心概念 展示详细信息、供阅读参考
空间利用 大字+留白营造高级感 内容填满但排版整齐不拥挤
移动端优化 ✅ 字号已增大 35-40% 以适配移动端观看 ✅ 字号已增大 35-40% 以适配移动端观看

多图拆分规则

当内容在单张 1200x1800px 海报中放不下时,必须拆成多张图

  • 每张图都有独立的完整结构(标签+标题+内容+分隔线)
  • 多张图用序号区分:poster-01.pngposter-02.png ...
  • 拆分点以逻辑单元为准(如每个工具一张、每段 Prompt 一张)
  • 向用户说明:"内容较多,已生成 N 张海报"

工作流程

第一步:理解需求

向用户确认以下信息(如果用户未提供,主动询问):

  1. 主题内容 — 海报要表达的核心主题是什么?
  2. 模板类型
    • 封面-1 (cover-center): 居中图标型 — 合集/总览类
    • 封面-2 (cover-grid): 特性网格型 — 多要点/步骤类
    • 封面-3 (cover-list): 列表展示型 — 清单/教程类
    • 内页-1 (inner-tool): 工具详情型 — 单工具介绍
    • 内页-2 (inner-prompt): 提示词模板型 — Prompt/AI指令
    • 内页-3 (inner-persona): 人设系统型 — 角色设定
  3. 主题色(可选)— 不指定则智能匹配
  4. 具体内容 — 文字、列表、代码等

如果用户只给了模糊主题,智能推断并生成内容。

第二步:构建 HTML

阅读 references/design-system.md 获取完整规范。

生成要求:

  1. 严格遵循设计规范中的数值
  2. 画布: 1200x1800px,viewport meta 固定宽度
  3. 字体: 标题用思源宋体 900 weight,正文用思源黑体
  4. 图标: 内联 SVG
  5. 渐变: CSS linear-gradient

写入临时文件:/tmp/design-ads-output.html

第三步:渲染为 PNG

node ~/.claude/skills/design-ads/scripts/render.js /tmp/design-ads-output.html \x3C输出路径>.png

多张图时依次渲染。

自动复制功能:渲染完成后,HTML 源文件和 PNG 图片会自动复制到下载目录下的时间戳文件夹中:

~/Downloads/design-ads_2026-04-03_10-55-13/
├── design-ads-output-cover.html
├── poster-cover.png
├── design-ads-output-01.html
└── poster-01.png

第四步:交付结果

  1. 用 Read 工具打开图片让用户预览
  2. 说明生成了几张图、每张的内容
  3. 询问是否需要调整

内容生成指南

封面类(精简信息 + 大字撑满空间)

信息量控制:最多 3 个元素

  • 主标题(1-2 行,超大字号 160-175px)
  • 副描述(1 行或短段落,32-36px)
  • 图标 / 网格卡片 / 列表项(三选一)

封面排版要点:

  • 主标题字号要大到占据画布纵向 35-45% 的视觉比重
  • 元素之间用充足留白分隔(不是空白,是有意的呼吸空间)
  • 整体垂直居中偏上分布,底部留 10-15% 给分隔线
  • 移动端优化:所有字号已增大 35-40%,确保在手机上清晰可读

内页类(内容充实 + 统一正文 + 标题突出)

信息量:尽可能详细

  • 工具名/章节名(超大标题 95-110px)
  • 功能描述(2-4 句话,28-32px)
  • 特性列表(3-6 项,每项可含简短说明)
  • 代码/Prompt 块(如有,保留完整格式 26-28px)
  • 适用场景(2-4 条)

内页排版要点:

  • 标题区(标签+主标题+副标题):占上部 25-30%
  • 内容区(卡片/列表/代码块):占中部 50-60%,这是重点,要充实
  • 底部:分隔线固定在距底 150px
  • 正文统一用思源黑体 28-32px,不要混用多种正文字号
  • 标题与正文的字号比至少 3:1 以上
  • 移动端优化:所有字号已增大 35-40%,确保在手机上清晰可读

快速参考

模板 类型 信息量 关键元素
cover-center 封面 居中图标 + 超大标题(160-175px)
cover-grid 封面 Badge + 大标题(135-150px) + 2x2网格
cover-list 封面 Badge + 大标题(135-150px) + 列表卡片
inner-tool 内页 充实 标题(95-110px) + 描述 + GitHub卡
inner-prompt 内页 充实 标题(95-110px) + 代码块 + 标签
inner-persona 内页 充实 标题(95-110px) + 系统提示词 + 场景列表

移动端优化:所有字号已增大 35-40%,专为移动端观看优化

查看 assets/ 目录下的 6 张模板图片作为视觉参考。

Usage Guidance
What to consider before installing: - The skill is coherent: it generates HTML per the included design spec and uses Puppeteer/Selenium to render PNGs. No secrets or network exfiltration were found. - It will run local code (Node scripts and optional Python Selenium) and requires Node.js >= 18 and a Chrome/Chromium binary. Installing Puppeteer/npm deps may download a large Chromium binary if system Chrome is not used. - The render script will create files in /tmp and copy the HTML + PNG outputs into a timestamped folder under your Downloads directory and will enumerate folders there (to find recent design-ads_* dirs). If you do not want local files written or directory reads, do not install or run the skill. - The script launches Chrome with flags like --no-sandbox (common for headless runs) and sets an executablePath hardcoded to a macOS Chrome path; on non-macOS systems you may need to adjust configuration or ensure a Chromium binary is available. - As with any code from an unknown source, review the bundled scripts before executing and run in a controlled environment if you have concerns (e.g., sandboxed VM or container). If you need additional assurance, ask the publisher for provenance or sign-off, or request a version that omits automatic copying to ~/Downloads.
Capability Analysis
Type: OpenClaw Skill Name: design-ads Version: 1.0.0 The skill bundle contains a script (`scripts/render.js`) that accesses the user's home directory (`os.homedir()`) and automatically copies generated HTML and PNG files to the `~/Downloads` folder. While this behavior is explicitly documented in `SKILL.md` as a feature for user convenience, it constitutes a high-privilege file system interaction outside the standard agent workspace. Additionally, the rendering scripts in the `scripts/` directory utilize Puppeteer and Selenium with the `--no-sandbox` flag, which increases the security risk during the processing of generated HTML content.
Capability Assessment
Purpose & Capability
Name/description (poster/cover generator) aligns with the included scripts (render.js, puppeteer renderers), design-system reference, and Puppeteer dependency — these are expected for rendering 1200x1800 PNGs from HTML.
Instruction Scope
SKILL.md instructs the agent to generate HTML, save to /tmp/design-ads-output.html, then run the bundled render script which loads the local HTML via file://, screenshots using Puppeteer/Selenium, and presents the PNG. The runtime will read the local references/design-system.md and write files to /tmp and the user's Downloads folder. This file I/O is consistent with the skill purpose but users should expect local file creation and directory enumeration of ~/Downloads.
Install Mechanism
There is no separate download-from-URL installer; package.json lists Puppeteer (a standard npm dependency) and code is included in the bundle. No external or shortener URLs or extract-from-untrusted-archive behavior was found.
Credentials
The skill requires no environment variables or credentials. It does access the filesystem (home and Downloads directories) to create timestamped output folders and to copy generated files, which is proportionate to delivering output but worth noting to users who may not want files written or directories enumerated.
Persistence & Privilege
always is false and the skill does not request persistent platform-wide privileges or modify other skills. It only writes its own output files to /tmp and ~/Downloads and does not alter agent configuration.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install design-ads
  3. After installation, invoke the skill by name or use /design-ads
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of the "design-ads" skill: an AI-powered vertical poster/cover image generator for knowledge topics. - Automatically creates 1200x1800 px posters in a unified dark “tech” style, using Source Han Serif and Sans fonts, gradient accents, and six layout templates. - Accurately distributes content inside a 960x1500 px area, following strict space and typography rules for covers (large titles, minimal info) and inner pages (detailed layouts). - Supports multi-image splitting for longer content, mobile-friendly font scaling, and guides users to clarify topic, template type, theme color, and content. - HTML output is auto-rendered to PNG via Node/Puppeteer, with files copied to a timestamped downloads folder for easy access.
Metadata
Slug design-ads
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is design-ads 海报制作?

AI 知识类海报/封面图设计生成器。根据用户提供的主题内容,自动生成 1200x1800 像素的竖版海报图片, 风格统一为深色科技感设计(暗黑背景 + 渐变强调色 + 思源宋体标题 + 思源黑体正文)。 支持 6 种模板布局:3 种封面样式(居中图标型、特性网格型、列表展示型)和 3 种内页样式(工具详情型、提示... It is an AI Agent Skill for Claude Code / OpenClaw, with 108 downloads so far.

How do I install design-ads 海报制作?

Run "/install design-ads" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is design-ads 海报制作 free?

Yes, design-ads 海报制作 is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does design-ads 海报制作 support?

design-ads 海报制作 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created design-ads 海报制作?

It is built and maintained by bozoyan (@bozoyan); the current version is v1.0.0.

💬 Comments