← Back to Skills Marketplace
panhongwei

html-report

by panhongwei · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
347
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install html-report-generator
Description
将任意输入自动分解成 5-15 个独立 HTML 报告页面,每页严格 1017×720px(对齐 PPT 画布 10.59"×7.499" @96dpi),深度拆解 3-6 个子维度,每维度精炼 60-100 字 + 支撑图表。当用户说"生成报告"、"分析内容做成页面"、"做成HTML"、"内容可视化"时立即使用...
README (SKILL.md)

HTML 报告生成器 · 主索引

本 skill 拆分为 10 个专项文件,按需读取,不要预加载所有文件


📂 文件索引

文件 职责 生成时机
01-canvas.md 画布尺寸、四区结构、溢出规则 每页开始前
02-design-system.md 自创 Tc 模板流程、6种基础模板基因库、主题一致性 规划阶段(第零步)
03-layout.md 自创 Lc 布局流程、4种基础布局基因库、空间计算 每页选布局时(第零步)
04-color-font.md 7套配色、字体规则、语义色系统 每页设定样式时
05-content.md 反偷懒约束、内容密度、基础SVG图表库(6种) 写内容时
06-workflow.md 主题拆解规划、渲染验证、质量核查 开始和结束时
07-special-pages.md 封面/目录/章节分隔/结尾页规范与HTML模板 生成特殊页面时
08-svg-extended.md 扩展SVG图表库(圆环/散点/流程/树状/热力/瀑布/双轴) 需要复杂图表时(可选)
09-components.md 页眉/摘要栏/卡片变体/徽章/图标/数字排版规范 每页组件选用时
10-diagram-types.md 业务图谱库:架构图/流程图/层级图/决策图 + 网络查询流程 用户要求汇报图谱时

⚡ 三条不可逾越的铁律

违反任意一条 → 该页推翻重做,无例外。

铁律 1:画布锁死 1017×720px

*, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; }
html, body {
  width: 1017px;  height: 720px;
  min-width: 1017px; max-width: 1017px;
  min-height: 720px; max-height: 720px;
  overflow: hidden;
}

铁律 2:四区高度必须精确求和 720px

Header   72px  ←  页眉
Content 580px  ←  主内容(仅左右 padding,上下为 0)
Summary  48px  ←  摘要栏(必须有实质内容,不能留空)
Footer   20px  ←  页脚
─────────────
总计    720px  ✓

铁律 3:不得使用 LibreOffice 渲染

必须使用 Chrome/Puppeteer 截图(见 06-workflow.md)。


🚀 完整生成流程

阶段一:规划(生成任何页面前)

1. 读 02-design-system.md → 第零步:分析用户描述,自创 Tc 专属模板
2. 读 03-layout.md        → 了解 Lc 自创布局机制
3. 读 06-workflow.md      → 拆解主题(8-12维度),规划每页三元组
4. 读 07-special-pages.md → 确认是否需要封面/目录/章节/结尾页

阶段二:逐页生成

每页生成时:
├─ 读 01-canvas.md        → 确定四区 CSS 骨架
├─ 读 02-design-system.md → 应用 Tc 模板的背景/卡片/页眉 CSS
├─ 读 03-layout.md        → 第零步:自创本页 Lc 布局
├─ 读 04-color-font.md    → 选配色方案 + 字体
├─ 读 05-content.md       → 填内容,选基础SVG图表,反偷懒自查
├─ 读 08-svg-extended.md  → 需要圆环/流程/树状等复杂图表时(可选)
├─ 读 09-components.md    → 选择页眉变体/摘要栏/卡片变体/徽章
└─ 读 10-diagram-types.md → 用户要求架构图/流程图/层级图/决策图等业务图谱时(可选)
                            若图谱不在库中 → 执行文件末尾的 WebSearch 查询流程

特殊页面(封面/目录/章节页):
└─ 读 07-special-pages.md → 使用对应变体模板

阶段三:验证与收尾

└─ 读 06-workflow.md → 运行截图验证,对照质量清单

📋 报告结构规范

P00  封面页(Cover)       → 07-special-pages.md CV变体
P01  目录页(可选)        → 07-special-pages.md TC变体
──── 章节分隔(可选)      → 07-special-pages.md CD变体
P01+ 内容页 × N           → 主流程
──── 章节分隔(可选)      → 07-special-pages.md CD变体
PN   结尾页(End)         → 07-special-pages.md EP变体

🎯 图表选型速查

数据特征 图表类型 文件
趋势/时间序列 折线面积图 05 图表2
量级对比(同类) 横条图/柱状图 05 图表1/6
多维度评分 雷达图 05 图表3
时间节点 时间轴 05 图表4
进度/计划 甘特条 05 图表5
占比/构成 圆环图 08 图表7
相关性分析 散点图 08 图表8
步骤/决策流 流程图 08 图表9
层级/树状 树状图 08 图表10
二维密度 热力矩阵 08 图表11
累计变化 瀑布图 08 图表12
双量级叠加 双轴图 08 图表13
系统组件/微服务/网络拓扑 分层架构图 / C4上下文图 10 一
跨部门流程/审批/交接 泳道图 10 二
转化率/用户路径 漏斗图 10 二
流量/预算分配流向 Sankey桑基图 10 三
KPI目标对比 Bullet子弹图 10 三
市场占比/资产组合 Treemap树图 10 三
组织/优先级/市场矩阵 金字塔/2×2矩阵/同心圆 10 四
风险评估/任务分配 风险矩阵/RACI 10 五
不在以上列表的图谱 → 读 10-diagram-types.md 末尾执行 WebSearch 10 末
Usage Guidance
This skill largely looks like a design-and-template-driven HTML report generator, but there are a few mismatches you should resolve before enabling it for sensitive use: - Verify runtime requirements: SKILL.md expects Chrome (Puppeteer) and Node scripts (screenshot_batch.js, present_files). Make sure those binaries/tools exist and are safe to run in your environment — the skill's manifest does not declare or install them. - Confirm network behavior: the references mention running a "WebSearch" when a requested diagram is not in the library. Ask the maintainer (or inspect the agent's runtime bindings) what endpoint/service the agent will call and whether report content (potentially sensitive) will be sent externally. - Check file outputs and permissions: the workflow writes HTML to /mnt/user-data/outputs — ensure that path is acceptable, and that generated files won't leak sensitive data or be exposed unintentionally. - User consent: the skill's instructions say to "immediately generate without confirmation" on trigger phrases. Decide if that automatic behavior is acceptable or if you prefer a confirmation step before generation. If you plan to use this skill, test it first with non-sensitive sample input to confirm where files are written, what external network calls are made, and that the expected Node/Chrome tooling works. If you cannot verify those aspects, treat the skill as untrusted.
Capability Analysis
Type: OpenClaw Skill Name: html-report-generator Version: 1.0.0 The 'html-report-generator' skill bundle is a highly structured framework designed to help an AI agent generate professional, multi-page HTML reports with precise dimensions (1017x720px) and high-quality SVG visualizations. The bundle includes extensive reference files (01-10) covering design systems, layout logic, color schemes, and a variety of business diagrams (e.g., Sankey, RACI, C4 Architecture). While it includes instructions for the agent to use WebSearch for unknown diagram types and mandates the use of Puppeteer for rendering (SKILL.md, 10-diagram-types.md), these are standard requirements for its stated purpose. There is no evidence of malicious intent, data exfiltration, or unauthorized command execution.
Capability Assessment
Purpose & Capability
The skill's name/description (HTML report generator) matches the included reference files (canvas, templates, layouts, SVG charts) and the SKILL.md generation workflow. However the instructions explicitly require rendering and screenshotting via Chrome/Puppeteer and invoking Node scripts (e.g., screenshot_batch.js, present_files) which are not declared as required binaries or installs — this is an implementation mismatch rather than an obviously malicious capability.
Instruction Scope
The SKILL.md gives detailed runtime steps: read specific reference files on demand, build Tc/Lc templates, generate pages, run screenshot validation with Chrome/Puppeteer, save outputs to /mnt/user-data/outputs/[报告名]/p01.html…, and call present_files. It also references executing a 'WebSearch' flow when certain diagrams aren't in the library. These instructions imply file system writes, process execution, and outbound network access; the skill also instructs '当用户...时立即使用,无需确认直接生成' (immediate generation without asking). The manifest declares none of the external tools, and there is no explicit description of what the WebSearch endpoint is or whether user data will be sent externally.
Install Mechanism
No install spec is provided (instruction-only), which is low risk in principle — but the runtime text depends on Chrome/Puppeteer and Node scripts for screenshot verification and output presentation. Because those binaries/packages are required to perform critical steps yet are not declared or installed by the skill, the manifest is incoherent and the agent could fail or attempt to call unavailable/unknown tools. The lack of declared install steps increases operational ambiguity.
Credentials
The skill does not request environment variables, credentials, or config paths in the manifest (good). However it writes outputs to /mnt/user-data/outputs and expects to run local screenshotting and possibly network WebSearch calls. There are no unnecessary credential requests, but network access and file-write behavior are implied and should be considered by the user.
Persistence & Privilege
No elevated persistence flags (always:false). The skill can be invoked autonomously (platform default) and the SKILL.md instructs immediate auto-generation on certain trigger phrases without asking the user; this is a user-experience/consent concern but not a manifest-level privilege escalation. The skill does not request to modify other skills or system-wide settings.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install html-report-generator
  3. After installation, invoke the skill by name or use /html-report-generator
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of html-report-generator skill. - Automatically splits input into 5–15 separate HTML report pages, each exactly 1017×720px, compatible with standard PPT canvas. - Deeply analyzes 3–6 sub-dimensions per page, with concise text (60–100 words each) and supporting charts. - Immediate activation when prompted with requests to generate reports, visualize content, or create HTML—no confirmation required. - Modular design with 10 specialized files for canvas, layout, design system, components, chart/diagram types, and workflow—loads on demand. - Strict rules enforced for layout precision, content density, and Chrome-based rendering only.
Metadata
Slug html-report-generator
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is html-report?

将任意输入自动分解成 5-15 个独立 HTML 报告页面,每页严格 1017×720px(对齐 PPT 画布 10.59"×7.499" @96dpi),深度拆解 3-6 个子维度,每维度精炼 60-100 字 + 支撑图表。当用户说"生成报告"、"分析内容做成页面"、"做成HTML"、"内容可视化"时立即使用... It is an AI Agent Skill for Claude Code / OpenClaw, with 347 downloads so far.

How do I install html-report?

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

Is html-report free?

Yes, html-report is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does html-report support?

html-report is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created html-report?

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

💬 Comments