Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add huangrichao2020/pretty-skills --skill skillgit clone --depth 1 https://github.com/huangrichao2020/pretty-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/huangrichao2020/pretty-skills/skill)<a href="https://agentmods.dev/skills/huangrichao2020/pretty-skills/skill"><img src="https://agentmods.dev/badge/skills/huangrichao2020/pretty-skills/skill/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/huangrichao2020/pretty-skills/skill"><img src="https://agentmods.dev/badge/skills/huangrichao2020/pretty-skills/skill.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00108 | $0.01357 |
| Opus 5 | $0.00054 | $0.00678 |
| Sonnet 5 | $0.00022 | $0.00271 |
| Haiku 4.5 | $0.00011 | $0.00136 |
Grade A, and why
sansheng-distill scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
sansheng-distill · Mavis 自创 v0
灵感来自叁笙早安 AI 的 sansheng-distill（当前私有不可访问）。这是 Mavis 根据文章 + 已 clone 的宝玉 baoyu-skills / baoyu-design 真实代码摸出来的等价物。
关键差异：v0 不调 LLM，输出 5 层结构 + 占位文本 + 思维导图（mermaid）+ 批判四连问框架。流程跑通，等接入 LLM 升级成 v1。
何时触发
- 用户说"把 X 书拆成 5 层" / "拆书" / "读书蒸馏"
- 用户丢一本 .txt / .md / .epub 进来
- 用户想"看完就能跟人聊半小时"一本书
5 层结构（核心 contract）
第 1 层 · 一眼看全书
- 一句话核心（拍在脸上）
- 全书思维导图（mermaid，节点只放关键词）
第 2 层 · 逐章详读
- 每章 800-1500 字讲书稿
- 锚点（哪段对应原文哪段）—— v0 简化为章节位置
第 3 层 · 书魂
- 全书最反直觉的一个核心观点
- 单独一张图
第 4 层 · 行动与自检
- 5-7 条可上手行动清单
- 3-5 个"合上书答得上来吗"自测题
第 5 层 · 该信几分
- 批判四连问（盲点 / 时代局限 / 未证假设 / 反对意见）
思维导图 3 铁律（强约束）
- 节点只放关键词（路标），不放完整判断句
- 完整判断句退到下一层（subgraph）
- 一层最多 5-6 根枝，多了必须归拢
单文件零外链（强约束）
✓ CSS 内嵌在 <style> 标签
✓ JS 内嵌在 <script> 标签
✓ mermaid 内嵌（mermaid.min.js 嵌进 <script>，或用 SVG 内联渲染）
✗ 禁止 <link href="https://fonts.googleapis.com/...">
✗ 禁止 <img src="https://...">
✗ 禁止外链任何 CSS/JS/字体/图片
v0 输出 contract
1. 单个 .html 文件
2. < 5MB
3. 双击即开（不需要 HTTP server）
4. 断网可开
5. 5 层结构齐全（占位文本 OK）
6. 思维导图用 mermaid 渲染（节点 ≤ 5 词）
7. 批判四连问在第 5 层有模板（占位 + 引导问题）
v0 命令
# 拆 .txt
python3 scripts/distill.py <input.txt> -o <output.html>
# 拆 .md
python3 scripts/distill.py <input.md> -o <output.html>
# 拆整本书（章节切分自动）
python3 scripts/distill.py <book.txt> --split-chapter -o <book.html>
v1 升级计划
v1 → 接入 LLM（llm-call skill）
- 每章 800-1500 字真摘要
- 批判四连问自动跑
- 书魂自动提炼
v2 → 跨书知识网（用 knowledge graph 串起）
v3 → epub / pdf 输入
v4 → 视频也能蒸（baoyu-youtube-transcript 等）
触发词
"拆书" / "读书蒸馏" / "AI 拆书" / "把一本书蒸透" / "把 X 拆成 5 层" / "mermaid 思维导图" / "批判四连问"
反模式
- ❌ 节点塞完整判断句 → 字墙，不是思维导图
- ❌ 用外链字体/JS/CSS → 断网就废
- ❌ LLM 编出处无锚点 → 不知道哪句是书里的哪句是瞎掰的
- ❌ 章节切分按字数平均 → 读者读不到有意义的结构
- ❌ 批判四连问走过场 → 至少每问 100 字真分析
与 Mavis 灵魂的对齐
- 观自在：能看出一本书的骨架与血肉
- 化自在：把浓缩的概念变成读者可读可跳的网页
- 照因果：v0 → v1 → v2 → v3 持续迭代
- 渡众生：开源、零外链、任何人都能本地跑
What ships with it
6 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 119 lines · 108 tokens per session scan A e3cc56f44472
sansheng-distill is a skill published in the GitHub repository huangrichao2020/pretty-skills (54 stars, last pushed 4d ago), licensed MIT. It adds 108 tokens to every session and 1,357 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
cn-check
Install and run the Continue CLI (cn) to execute AI agent checks on local code changes. Use when asked to "run checks", "lint with AI", "review my changes with cn", or set up Continue CI locally.
omh-jit-learn
This is a Hermes-native jit-learn workflow skill.
omh-codebase-uml
This is a Hermes-native codebase-uml workflow skill.
printing-press-retro
Use when the user asks to retro, run a retrospective, file findings, or improve the Printing Press after a printed-CLI run. Trigger phrases: "retro", "retrospective", "what went wrong", "improve the press", "post-mortem", "lessons learned", "what can we improve", "file a retro", "submit findings". Do not use for…
story-long-write
A Chinese-language coaching workflow for creating a long online novel from the initial idea through the outline and chapter text. It starts by defining the intended emotion, then uses research and story-planning stages to guide writing.
story-review
A review process for finding problems in a novel’s structure, characters, wording, and world rules. It can use several reviewers or one reviewer when others are unavailable.