← 返回 Skills 市场
nefas11

Maus HTML Summary

作者 Nefas11 · GitHub ↗ · v1.0.1 · MIT-0
cross-platform ✓ 安全检测通过
76
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install maus-html-summary
功能描述
Turn articles, transcripts, and raw text into a self-contained illustrated HTML explainer with simple language, memorable structure, and a "Sendung mit der M...
使用说明 (SKILL.md)

Maus HTML Summary

Turn a source text (article, transcript, talk, long post, raw text) into one self-contained HTML page that explains it with the clarity of a children's science show for adults: friendly, concrete, visual, source-faithful, and easy to remember.

The name references "Die Sendung mit der Maus", a German educational TV show known for explaining complex topics in simple visual steps. Outputs should be internationally understandable; do not assume the reader knows the show.

Goal

A single self-contained HTML page that is:

  • easy to scan
  • visually illustrated with real explanatory SVGs (not decorative icons)
  • plain-language but substantively concrete
  • faithful to the source and honest about what was left out
  • memorable through metaphor and clear hierarchy

The tone is "clear science explainer for adults": friendly, concrete, slightly playful, never childish, never vague.

The non-negotiable workflow

These seven stages run in order, every time. Skipping a stage produces shallow or hallucinated output. Reproducibility is the audit (Stage 7) catching what earlier stages missed.

Stage 1 - Internal extraction (no output yet)

Read the source. Build a short internal notes block. It never appears in the HTML; it is the only material allowed in the HTML.

Capture four buckets:

  • Facts - numbers, dates, names, places, technical terms
  • Mechanisms - how things work: concrete steps, cause -> effect, before/after
  • Examples - specific cases mentioned in the source
  • Source claims - what the source explicitly argues or shows

Anti-hallucination rule: if a claim is not in this list, it cannot appear in the HTML.

Stage 2 - Plan hierarchy

Count source words. Pick main-idea count by size:

  • \x3C 500 words (short transcript snippet, social post): 3 main ideas
  • 500-2000 words (article, talk excerpt): 4 main ideas
  • 2000+ words (long article, full transcript): 5-6 main ideas

For transcripts: first mentally filter filler words, false starts, hesitations, repetition. Extract from the cleaned signal, not the raw text.

For each main idea, plan 2-4 sub-points. Sub-points are where concreteness lives.

Identify the one-sentence core message; this becomes the hero.

Stage 3 - Pick an SVG metaphor per main idea

Every main idea carries exactly one explanatory SVG. Pick from this catalog:

  • Layer stack - layered concepts, abstraction levels, stacked components
  • Pipeline / conveyor belt - sequential processes, transformations, flows
  • Cross-section - what is inside something: anatomy, internals, structure
  • Before / after - change, transformation, improvement
  • Node graph - networks, relationships, dependencies, hubs
  • Funnel - filtering, narrowing, conversion
  • Growth curve - growth, decay, trends over time
  • Circuit / logic flow - logical wiring, if/then chains, branching flow

Use the skeleton SVGs in svg-patterns.md as the geometric base. Adapt labels, colors, element count to the specific content. Never copy-paste verbatim.

If none of the metaphors fit a given idea, the idea is probably too abstract to stand alone. Fold it into another idea.

Decorative emoji or icons are not SVG metaphors. A row of generic symbols does not satisfy the SVG requirement.

Stage 4 - Draft sections

Build the HTML around this arc:

  1. Hero / What is this about? - the one-sentence core message plus a 1-2 sentence lead.
  2. Main ideas section - 3-6 cards (per Stage 2), each with:
    • Headline
    • SVG metaphor (from Stage 3, adapted from svg-patterns.md)
    • 2-4 sub-points (each passing the concreteness check in Stage 5)
  3. Why it matters - one box, the broader significance as stated in the source.
  4. What we left out - honesty box: 2-4 things from the source that were intentionally omitted, each with a one-line reason ("too technical for this format", "side path not tied to the core message", "redundant with idea 2", ...).
  5. Memory hook - one memorable sentence, readable aloud in under 5 seconds.
  6. If you only remember 3 things - exactly three bullets, the absolute minimum take-away.

Stage 5 - Concreteness check per sub-point

Every sub-point must contain at least one of:

  • a number (count, percentage, duration, threshold, version)
  • a proper noun (product, person, company, technology, place)
  • a concrete example (a specific case from the source)
  • a mechanism step (how it works, not just that it works)

If a sub-point has none of these, it is too vague. Rewrite it using material from the Stage 1 extraction notes, or remove it.

Vague-by-default phrases that signal a failed sub-point:

  • "improves efficiency"
  • "plays an important role"
  • "enables new possibilities"
  • "is an important aspect"
  • "contributes to ..."
  • "is significant"

If you catch yourself writing any of these, you're describing the existence of a thing instead of explaining it. Go back to the extraction notes and pull a fact, name, example, or step.

Stage 6 - Render HTML

Write one complete, self-contained HTML document:

  • All CSS inline in \x3Cstyle>
  • All SVGs inline
  • No external scripts, fonts, images, fetches, or \x3Clink> tags
  • Mobile-friendly (verify mental layout at 360 px width)
  • Strong visual hierarchy: hero -> idea cards -> significance -> honesty -> memorable closing
  • Rounded shapes, warm muted palette, soft shadows, light borders
  • Hierarchy through size, weight, and color; not through heavy lines

Stage 7 - Pre-emit audit (the reproducibility gate)

Before delivering, walk this checklist mentally. Any failed item must be fixed before emitting.

[ ] Every sub-point contains a number, proper noun, concrete example,
    or mechanism step (Stage 5 rule)
[ ] Every main idea has exactly one explanatory SVG drawn from the
    catalog, adapted from svg-patterns.md; not a decorative icon
[ ] The "What we left out" box names at least 2 specific things from the
    source that were left out, each with a one-line reason
[ ] No fact in the output is missing from the Stage 1 extraction notes
    (no hallucinations)
[ ] The memory hook reads aloud in under 5 seconds
[ ] The "If you only remember 3 things" box has exactly 3 items
[ ] HTML is self-contained: no \x3Cscript>, no \x3Clink rel="stylesheet">,
    no \x3Cimg src="http...">, no external font URLs

The audit is the difference between consistent and inconsistent output. Don't skip it. Don't treat it as a wish list. It is a checklist.

Writing rules

  • Use the source language by default. If the user or audience context indicates English, use English.
  • Concreteness over brevity. If forced to choose between "short and vague" and "two sentences and concrete", pick concrete.
  • Explain jargon once, then drop it.
  • One narrative thread per main idea; don't mix topics inside one card.
  • Short paragraphs, but use lists where lists fit naturally (steps, ingredients, options).
  • Avoid long quotations. Summarize.

HTML rules

  • One complete document, self-contained, no external dependencies.
  • Mobile-first layout, generous padding.
  • Warm muted palette, but keep it globally readable and not region-specific.
  • Rounded corners, soft shadows.
  • Hierarchy via size, weight, color; not via heavy lines or borders.

When the source is unclear

If the source is genuinely ambiguous about a point, say so in the output inside the relevant card or in the "What we left out" box. Never invent details to fill space.

Resources

  • svg-patterns.md - skeleton SVGs for the eight metaphor types in Stage 3. Use as geometric base, adapt content.
安全使用建议
Before installing, be aware that the skill may be invoked implicitly when you ask for an illustrated HTML summary. Avoid using it on private or sensitive source text unless you are comfortable having your agent transform that content into an HTML document.
能力标签
crypto
能力评估
Purpose & Capability
The artifacts coherently support the stated purpose: summarize articles, transcripts, long posts, and raw text into source-grounded HTML with inline SVG explanations.
Instruction Scope
The OpenAI agent config allows implicit invocation for a broad text-to-HTML transformation, which could surprise users in multi-skill setups, but the behavior is disclosed and aligned with the skill's purpose.
Install Mechanism
The package contains markdown and YAML files only; there are no executable scripts, dependencies, install hooks, or package registry concerns.
Credentials
The skill instructs generation of self-contained HTML and explicitly forbids external scripts, fonts, images, fetches, and stylesheet links, so requested environment access is proportionate.
Persistence & Privilege
No persistence, background workers, credential access, privilege escalation, local indexing, or mutation authority appears in the artifacts.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install maus-html-summary
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /maus-html-summary 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.1
Internationalize skill description and SVG pattern labels; reduce German-specific wording while keeping the Maus inspiration clear.
v1.0.0
Initial release: illustrated HTML summaries with extraction, SVG metaphors, concreteness audit.
元数据
Slug maus-html-summary
版本 1.0.1
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 2
常见问题

Maus HTML Summary 是什么?

Turn articles, transcripts, and raw text into a self-contained illustrated HTML explainer with simple language, memorable structure, and a "Sendung mit der M... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 76 次。

如何安装 Maus HTML Summary?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install maus-html-summary」即可一键安装,无需额外配置。

Maus HTML Summary 是免费的吗?

是的,Maus HTML Summary 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Maus HTML Summary 支持哪些平台?

Maus HTML Summary 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Maus HTML Summary?

由 Nefas11(@nefas11)开发并维护,当前版本 v1.0.1。

💬 留言讨论