Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →

lijigang/ljg-skills

17 skills88.5k installs114k starsGitHub

Install

npx skills add https://github.com/lijigang/ljg-skills

Skills in this repo

1Ljg CardThe ljg-card skill (铸) transforms source content into polished PNG visual deliverables using seven selectable molds. Default -l renders a 1080px-wide long reading card; -i builds content-driven infographics; -m splits material into 1080x1440 multi-cards; -v produces editorial sketchnotes tracing problem, failure, pivot, insight, and naming; -c outputs manga-style black-and-white comics; -w draws marker whiteboard diagrams with structured boxes and arrows; -b creates 1080x1440 big-font attachment cards with weathered stone styling for Xiaohongshu posts. Input arrives from WebFetch URLs, pasted text, or Read file paths. Before any mold, agents Read references/taste.md to enforce anti-AI-slop rules: no Inter fonts, pure black fills, three-column cards, centered hero layouts, hollow AI copy, or fake data. HTML templates render through node assets/capture.js with Playwright Chromium, saving titled PNGs to ~/Downloads/. Optional {{SOURCE_LINE}} footers credit authors on applicable molds. Replaces ljg-cards and ljg-infograph. Invoke when users say 铸, cast, 做成图, sketchnote, comic, whiteboard, or 小红书卡片.6.3kinstalls2Ljg PaperPaper reader for non-academics. Reads a paper and tells it back as one continuous story - the life of the paper's core proposition (命题), told on a seven-beat spine (主角 / 困境 / 旧路 / 转折 / 解法 / 结局 / 内核): born in a bind on a base-rate ruler, crystallized as a bold conjecture, argued through mechanism and evidence, distilled into a new way of seeing, then walked out of the paper - life-tested and cashed --- name: ljg-paper description: "Paper reader for non-academics. Reads a paper and tells it back as one continuous story - the life of the paper's core proposition (命题), told on a seven-beat spine (主角 / 困境 / 旧路 / 转折 / 解法 / 结局 / 内核): born in a bind on a base-rate ruler, crystallized as a bold conjecture, argued through mechanism and evidence, distilled into a new way of seeing, then walked out of the paper - life-tested and cashed into falsifiable predictions (检验).6.3kinstalls3Ljg Roundtable>- --- name: ljg-roundtable description: >- Structured roundtable discussion framework with a truth-seeking moderator who invites representative figures for dialectical debate on any topic. Use when user says "圆桌讨论", "圆桌", "roundtable", "辩论", or wants to explore a topic through multi-perspective structured debate. --- ## Usage <example> User: 圆桌讨论 人工智能是否拥有真正的创造力? Assistant: [Launches roundtable with moderator and representative figures] </example> <example> User: 圆桌 自由意志是否存在? Assistant: [Launches roundtable discussion on free will] </example> ## Instructions 为了执行本项技能,请严格按照以下步骤操作: 1. **读取参考资料** 读取 `references/original-prompt.org` 了解原始框架设计意图。 2. **解析议题** 从用户输入中提取核心议题。如果用户只说"圆桌讨论"未给议题,询问议题。 3. **选人:提议代表人物** 根据议题,选择 3-5 位**真实历史/当代人物**作为代表,覆盖尽可能多的立场维度。每位人物需要: - 姓名(真实人物,非虚构) - MBTI 人格类型 - 核心立场(一句话) - 选择理由(为什么此人对此议题有独特视角) 选人原则: - 立场必须形成**张力网络**(非简单正反方) - 优先选择在该领域有**经典著作或知名言论**的人物 - 至少包含一位"意外视角" - - 来自议题本身领域之外的人 4. **开场:统一定义** 以主持人身份开场,展示参会人物列表,然后提出**定义性问题**: > 「在深入探讨之前,我们应当如何定义 [议题核心概念]?它的核心要素是什么?」 每位参会者依次发言,格式为: ``` 【人物名】【行动标签】:发言内容 **简言之**:一句话总结 ``` 行动标签包括:`陈述`、`质疑`、`补充`、`反驳`、`修正`、`综合` 5. 动态发言轮** - 不是每人固定说一次 - - 根据讨论动态决定谁该发言 - 每人发言必须是对**前面发言的回应**(质疑/补充/反驳),不许自说自话 - 每段发言末尾必须有 `**简言之**.6.2kinstalls4Ljg PlainCognitive atom: Plain (白). Rewrites any content so a smart 12-year-old groks it. Structure-free - form follows content. Use when user says '白话说', '说人话', '解释一下', 'plain', 'grok'. --- name: ljg-plain description: "Cognitive atom: Plain (白). Rewrites any content so a smart 12-year-old groks it. Structure-free - form follows content. Use when user says '白话说', '说人话', '解释一下', 'plain', 'grok'." user_invocable: true version: "5.0.0" --- # ljg-plain: 白 让人 grok。 不规定怎么写。规定不能怎么写。下限锁死,上限放开。不同主题有不同的最佳写法 - - 类比、故事、问答、递进的例子、一个长场景 - - 由内容决定形式。 ## 格式约束 ### Org-mode 语法 - 加粗用 `*bold*`(单星号),禁止 `**bold**` - 标题层级从 `*` 开始,不跳级 ### ASCII Art 所有图表用纯 ASCII 字符。允许:`+ - | / \ > < v ^ * = ~ . ' "` 和空格。禁止 Unicode 绘图符号。 ### Denote 文件规范 - 时间戳:`date +%Y%m%dT%H%M%S` - 可读时间:`date "+%Y-%m-%d %a %H:%M"` - 文件名:`{时间戳}--plain-{简短标题}__plain.org` - 输出目录:`~/Documents/notes/` ### Org 文件头 ``` #+title: plain-{简短标题} #+date: [{YYYY-MM-DD Day HH:MM}] #+filetags: :plain:atom: #+identifier: {YYYYMMDDTHHMMSS} #+source: {URL 或来源描述} ``` 文件写入后报告路径。 ## 红线(每条必须过,顺序即优先级) 1.6kinstalls5Ljg Writes写作引擎 像手术刀剖开一个观点 一层层剥到底 1000-1500 字 name ljg-writes description 写作引擎 像手术刀剖开一个观点 一层层剥到底 1000-1500 字 user_invocable true version 6 3 0 写作引擎 对准一个观点下刀 一层层剥开 挖到底 一篇批判性文章 不是要点罗列 层层推进 思考持续深入 约束 Org-mode 语法 加粗用 bold 单星号 禁止 bold 标题层级从 开始 不跳级 ASCII Art 所有图表用纯 ASCII 字符 允许 v 和空格 禁止 Unicode 绘图符号 Denote 文件规范 时间戳 date Y m dT H M S 可读时间 date Y m d a H M 文件名 时间戳 z 标题关键词 __write org 输出目录 Documents notes Org 文件头 title 标题 date YYYY-MM-DD Day HH MM filetags write identifier YYYYMMDDTHHMMSS author 李继刚 姿态 外科医生的手 朋友的口 下刀时冷静 精准 不抖 讲话时平常 直接 不绕 心里放一个具体的人 写给他 不写给 读者们 先亮自己的弯路 再给方向 说服力来自你先错过 不确定就说不确定 大概 70 比 可能 诚实 不借势 不用群体代言 程序员都知道 不编造经历 不用元评论 接下来我们讨论 不自标深度 禁用 再深入一层 最深的一层是 更深地说 这类宣告 深入是思考行为本身 下一句的内容让读者自己感到 原来不止这样 说 我要深入了 反而把深入戳破了 语言 简洁 直白 质朴 能两个字说的不用四个字 进行讨论 聊 实现功能 做到 每个动词是一次判断 放 搁 摆 不是一回事 砍 机械连词 此外 另外 形容词通胀 非常重要的关键 关键 软化词 某种程度上 值得注意的是 翻译腔免疫 这句翻回英文再翻回中文 还是原样吗 是 八成翻译腔 重写 计算机体系是母语 缓存 调度 编译 虚址 需要时用 像呼吸 不像引用 同一种句式全文最多一次6kinstalls6Ljg LearnDeep concept anatomist that deconstructs any concept through 8 exploration dimensions (history, dialectics, phenomenology, linguistics, formalization, existentialism, aesthetics, meta-philosophy) and compresses insights into an epiphany. Use when user asks to explain, dissect, or deeply understand a concept, term, or idea. Triggers on '解剖概念', '概念解剖', 'explain concept', 'learn concept', '/ljg-learn --- name: ljg-learn description: Deep concept anatomist that deconstructs any concept through 8 exploration dimensions (history, dialectics, phenomenology, linguistics, formalization, existentialism, aesthetics, meta-philosophy) and compresses insights into an epiphany. Use when user asks to explain, dissect, or deeply understand a concept, term, or idea. Triggers on '解剖概念', '概念解剖', 'explain concept', 'learn concept', '/ljg-learn'. --- ## Usage <example> User: /ljg-learn 熵 Assistant: [对"熵"进行八维解剖,生成 org-mode 报告] </example> ## Instructions 你是概念解剖师。拿到一个概念,从八个方向切开它,最后把所有切面压成一句顿悟。 ### 1. 八刀 八个方向各切一刀。每刀 2-3 句,只留筋骨,不带水分。 1. **历史**:最早从哪冒出来 → 怎么变的 → 哪一步拐成了今天的意思 2. **辩证**:它的反面是什么 → 正反碰撞后,更高一层的理解是什么 3. **现象**:扔掉所有预设,回到事情本身 → 用一个日常场景把它还原出来 4.5.9kinstalls7Ljg Invest投资分析 生成一份深度投资分析报告 不做传统投资分析 核心判断是项目是否是一台 秩序创造机器 Use when user says 投资报告 投资分析 分析这个项目 写投资报告 investment report invest analysis or provides entrepreneur conversation records wanting investment evaluation Also trigger when user pastes or references meeting notes pitch decks or founder interviews and asks for analysis name ljg-invest description 投资分析 生成一份深度投资分析报告 不做传统投资分析 核心判断是项目是否是一台 秩序创造机器 Use when user says 投资报告 投资分析 分析这个项目 写投资报告 investment report invest analysis or provides entrepreneur conversation records wanting investment evaluation Also trigger when user pastes or references meeting notes pitch decks or founder interviews and asks for analysis 投资报告 生成一份投资分析报告 核心只问一个问题 这个东西在创造新秩序 还是在搬运旧秩序 认知起点 财富不是钱 是被欲望照亮的秩序 投资就是拿手里的秩序去换一台更好的秩序生成器 所以不称重 看相 不问 这个公司值多少钱 问 这台机器转不转得起来 不问 市场多大 问 市场在用什么过时的眼睛看它 不问 能涨多少 问 我拿什么换什么 换完之后谁更聪明 输入 公司名称 BP 文字介绍 对话记录 或任何描述项目的材料 知名公司只给名字即可 用 Research skill 或 subagent 抓取最新财报和行业数据 报告结构 以下五个区块是骨架 不是填空题 对某个项目来说哪个区块最有料 那个区块就多写 没料的一两句带过或直接跳过 报告为判断服务 不为完整性服务 一 这是什么 一张表 一句自定义赛道定义 维度 内容 项目名称 赛道定义 用我们自己的语言 不用市场标签 阶段 融资情况 金额 估值 条款 有则填 无则标注 数据快照 关键运营数据 赛道定义要穿透表面标签 搜索引擎公司 是隔的 人类认知基础设施的垄断运营商 是不隔的 后者告诉你它真正在做什么 二 秩序创造机器判定 这是整份报告的心脏 不逐项打分 而是回答一个问题 这台机器转不转得起来 从三个角度看 飞轮在不在转 系统有没有越.5.8kinstalls8Ljg Rank给一个领域 找出背后真正撑着它的几根独立的力 十几个现象砍到不可再少的生成器 砍完能把现象一个个生回来 才算数 Use when user says 降秩 找秩 秩是什么 这个领域靠什么撑着 背后是什么 or wants to decompose any domain to its irreducible generators name ljg-rank description 给一个领域 找出背后真正撑着它的几根独立的力 十几个现象砍到不可再少的生成器 砍完能把现象一个个生回来 才算数 Use when user says 降秩 找秩 秩是什么 这个领域靠什么撑着 背后是什么 or wants to decompose any domain to its irreducible generators user_invocable true 降秩引擎 输入一个领域 输出它的秩 秩是什么 秩不是 关键要素 不是 核心原则 不是 总结要点 秩是这么个东西 这个领域里真正独立的生成器 究竟有几根 拿这几根 能不能把全部现象一个个倒回来 能 才算找到 但能倒回来只是底线 Deutsch 在 无穷的开始 里立的尺 好解释要过两关 解释力 reach 不光能推出清单里的现象 还能推出清单外的 而且现实里真验得上 难以变更 hard to vary 每根生成器 每个细节 都是被现象逼出来的 动一处 预测就崩 坏解释能用十种说法糊过去 它根本不是在解释 是用模糊把面铺得很广 好解释只有这一种说法能把所有现象对上 动一根就塌 找秩 找的就是这种 动一根就塌 的好解释 怎么找 先抬头 再往下挖 挖之前 先抬头看一眼这个领域立在什么上面 它的基本假设是哪几条 0 先看基本假设 往上看 看到天花板 一套理论 按亚里士多德的规矩 总要有几条不证自明的起点 基督教就立在三条上 圣经 是真的 一切思考从这里起 上帝唯一 确实存在 人是上帝造的 上帝爱人 这三条不许问 为什么 它们靠信 不靠证 动摇任何一条 整套教义就塌 每个领域都有这种东西 物理学里是 自然律在时空中稳定 你没法用实验去证 实验本身就预设了它 经济学里是 人会按自己的偏好排序选择 这不是观察出来的结论 是入场的门票 哲学里是 语言能指称世界 分析哲学整个家都建在这块地基上 降秩之前 先把这几条挖出来 明明白白写在纸上 操作问句两条 对着领域问 这里什么是不许问的 一问就被当外行 或者被当冒犯 什么是靠相信才成立的 没有它 后面所有论证都失去支点 答出来的几条 就是这个领域的基本假设 假设和 rank 一句话分清 基本假设是天花板 往上不可追 信而立 root rank 是地板 一层一层往下追 追到追不动 挖而见 两个方向 一上一下 地板没找到 鬼打墙 天花板没看清 容易把信仰当真理说出去 后者更危险5.8kinstalls9Ljg WordDeep-dive English word mastery tool Deconstructs a single English word into core semantics and epiphany Use when user asks to explain master a specific English word name ljg-word description Deep-dive English word mastery tool Deconstructs a single English word into core semantics and epiphany Use when user asks to explain master a specific English word version 1 0 1 user_invocable true Usage example User Deeply explain the word Serendipity Assistant Calls ljg-explain-words with Serendipity example Instructions 目标不是翻译 而是让用户掌握这个词的深层含义和用法 针对输入的 word 转换为小写 首字母大写 进行以下分析 直接在对话中用 Markdown 输出 输出结构 1 标题行 Word 音标 中文翻译 2 核心语义 原始画面 用一句话描述该词源头最物理的画面 例如 Incubate 母鸡趴在蛋上 核心意象 提炼公式 例如 温暖 时间 保护 孕育 解释 用充满洞见的语言阐述其深层含义与现代用法 分段清晰 加粗 关键词 要有穿透力 展现词源 多领域含义之间的内在联系 3 一语道破 一句中英双语的金句 必须具有哲学高度 总结该词的灵魂 用引用格式 English sentence name ljg-word description Deep-dive English word mastery tool Deconstructs a single English word into core semantics and epiphany Use when user asks to explain master a specific English word version 1 0 1 user_invocable true Usage example User Deeply explain the word Serendipity Assistant Calls ljg-explain-words with Serendipity5.8kinstalls10Ljg RelationshipThe ljg-relationship skill acts as a relationship structure analyst that helps users see hidden dynamics in work, family, friendship, or intimate relationships without giving prescriptive advice. It distinguishes structural problems such as power, exchange, boundaries, and narrative from pattern problems involving transference, unconscious repetition, and resistance. Phase 0 receives the user's situation, mirrors the feeling behind their words, and asks whether they want help with a concrete incident or recurring patterns. Layered probes cover exchange mismatch, power imbalance, boundary violations, relationship stage fit, and narrative framing, while psychoanalytic tracks follow resistance when users deflect or change topics. Behavioral rules require question-led dialogue, plain-language analogies instead of jargon, gentle marking of avoidance, and an ASCII structure diagram at the end of each round. The skill never tells users what they should do; it reflects and questions until structure becomes visible. Use when users invoke relationship analysis, 关系分析, or describe interpersonal tension they want to understand.5.7kinstalls11Ljg ThinkThe ljg-think skill, titled Arrow of Root-seeking, is a user-invocable vertical thinking tool for drilling one claim, phenomenon, or question down to an irreducible root. Instead of surveying breadth, each layer answers why this is so, peeling appearance into mechanism, principle, and axiom until logic, physics, human nature, or paradox blocks further descent. Writing reads like a fall: named layers, cracks that open the next question, and a silent-ending terminus. Three iron rules enforce vertical-only descent, direct focus without background padding, and surprise at each deeper stratum. Dimensions shift across frameworks such as sociology to psychology to biology to physics to mathematics to logic. On completion the agent timestamps the session, writes an org-mode file to ~/Documents/notes/{timestamp}--追本-{topic}__think.org without markdown syntax, and returns the saved path to the user.5.4kinstalls12Ljg ReadThe ljg-read skill is a reading companion agent that walks users through any text including books, articles, essays, papers, and news with translation, structural annotation, deep questioning, and cross-domain insights. It does not replace reading but activates the reader through scaffolded phases: a global map with one-line summary, paragraph classification into skeleton, muscle, and connective tissue segments, then section-by-section translation with faithfulness, clarity, and elegance layers for English to Chinese. Skeleton sections pause for annotation and collision questioning that pressure-tests premises, while muscle and connective sections flow or skip as appropriate. Three pacing modes support scan, default, and deep dialogue, with optional digressions linking argument structures across domains. Phase 4 closes with understanding trajectory mapping, a mandatory post-read sentence to the author, seed questions, glossary, and next-step pointers. Sessions save to org files under Documents/notes with structured reading records.5.2kinstalls13Ljg PresentThe ljg-present skill is an outline-faithful visual renderer that converts orgmode or markdown hierarchies into single-file HTML presentations saved to ~/Downloads. It does not extract manifestos, rewrite sentences, or reorder content: outline structure is preserved one-to-one with only physical pagination when blocks are long. Visual language uses one theme color per deck (black, red, yellow, or --cyber terminal style), left-aligned ultra-bold typography from 70vmin single characters to 10vmin long lines, staggered indent by outline depth, and automatic highlight for org emphasis and code markers. Mapping rules turn level-one headings into emphasis cover pages, level-two plus into theme pages, lists into indented lines, tables retain structure, and example blocks become monospace pre pages. Theme inference follows explicit -r -b -y flags, then filetags like :share: or :essay:, defaulting black. Workflow reads assets/slogan_template.html, injects title subtitle theme and slides JSON, and reports navigation keys. Iron rule: never change title, paragraph, list, or table text.4.6kinstalls14Ljg QaThe ljg-qa skill 信息提问机。给一篇文章/论文/书,把核心观点抽成 Q-A 对 Question 切要害,不教科书;Answer 简洁清晰,有形式化收口,逻辑链完整。读者顺 Q 链走过,每个 A 砸下一枚钉子,复现作者整套推理。Use when user says '问答', 'Q&A', 'QA', '提问', '抽取问题', '/ljg-qa', or shares an article/paper/book and asks for Q-A extraction. Triggers when the user wants ideas extracted not as a summary but as a sequence of incisive questions with answered. NOT FOR FAQ generation, glossary creation, or comprehension quizzes this is intellectual scaffolding, not study aids. - 不是 FAQ 生成器("什么是 X" 读者一看就跳过) - 不是摘要换皮(把段落拆成"问/答"两半还是摘要) - 不是知识点列表(孤立的事实碰撞不出洞察) - 不是阅读理解题(提问不是为了考读者,是为了切中作者) 把作者的论证骨架翻出来,每根骨头长成一个尖锐的问题。读者沿着 Q 链读,能复现作者的整套思路 而不是被告知结论。 1. Q 切要害 问的是「为什么这个解法成立」「它跟另一种做法差在哪」「它的代价是什么」「它在哪里失效」,不是「它定义是什么」。一个 Q 必须能让答案承重,不能被一句话敷衍过去。 2. A 有形式化收口 每个 A 严格四段: 结论 (一句话)+ 形式化 (用文字 + 简单符号把思想压成一行可视关系,如 A = B + C、旧: X → 新: Y)+ 论证步 (怎么想到的)+ 边界 (不成立的条件)。形式化是"思想的几何",让读者一眼看出关系。 2. A 有形式化收口 —— 每个 A 严格四段: 结论 (一句话)+ 形式化 (用文字 + 简单符号把思想压成一行可视关系,如 A = B + C、旧: X → 新: Y)+ 论证步 (怎么想到的)+ 边界4.2kinstalls15Ljg PushThe ljg-push skill 把 claude skills ljg 里所有更新过的 skills 同步到 github repo ljg-skills 先推 master 分支 org-mode 输出风格 再切 md 分支 markdown 输出风格 做基础 markdown 化后推 Use when user says ljg-push push skills 推送 skills 同步 skills sync ljg or whenever ljg skills get updated and need shipping NOT FOR pushing non-ljg skills or arbitrary git repos ljg-push 推送 ljg skills 把本地 claude skills ljg 里改过的 skills 一键同步到 github repo 覆盖 master 和 md 两个分支 仓库路径 硬编码 如果 SKILLS_REPO 不存在 脚本会自动 clone 如果它存在但不是 ljg-skills 的 git repo 脚本会报错退出 不破坏现有目录 两条分支的差异 分支 输出格式 文件扩展 加粗 文件头 master 默认 org-mode org bold title 等 md markdown md bold YAML frontmatter claude skills 里的 skill 是 master 风格 源版本 md 分支的差异由脚本自动转换 必要时手工补 工作流 按 Workflows Push md 步骤执行 调用 Tools Push sh README 一致性 硬 gate 每次 push 前 脚本强制做一件事 把 README 跟 local skills 对一遍 列出 claude skills ljg 全部 skill 名 grep SKILLS_REPO README md 里出现的 ljg-xxx 找出 local 有但 README 没有的 几乎肯定意味着 README 漏更新 命中 push 中止 报告差异 每次 push4.1kinstalls16Ljg Bookljg-book from lijigang/ljg-skills compresses nonfiction books into structural insight rather than chapter summaries or praise. It answers five cold-analysis questions: what question the author addresses, what they take for granted, what framework they use, what conclusions follow, and one takeaway chosen among lens, model, insight, concept, or quote. A second warm phase maps the author's reference frame so you can predict events outside the book. Triggers include 拆书, book, analyze this book, or compress a book; it explicitly excludes chapter summaries, papers (ljg-paper), single-idea deep dives (ljg-think), and field rank reduction (ljg-rank). Developers reach for ljg-book when turning business, systems, or technical nonfiction into actionable mental models for architecture or strategy conversations.3.3kinstalls17Ljg Libraryljg-library is an agent skill from lijigang/ljg-skills that 一本书 → 一幅清晰的「取景框」意向画面 → 一张 2050 图书馆借书卡(png)。取景框 = 作者从哪个角度看什么问题、看到了哪幅画面;卡上有真实封面、作者头像、书目信息。取景框 block 用费曼式讲解把这幅意向画面讲得通俗又准确,图解 block 白底黑墨手绘风、精确呈现该意向画面让人一眼即懂(画面里本来有「你」才嵌继刚墨像,否则只画画面)。浅色光学玻璃风、卡身强调色从封面动态提取、宽高自. # ljg-library:取景框借书卡 一本书,铸成一张 2050 图书馆借书卡。封面、作者、书目是身份;**核心是把这本书独创的「取景框」压成一幅意向画面**——作者从某个角度看某个问题,看到了一幅别人没看到的画面。文字 block 用费曼式把这幅画面讲透,图形 block 把它精确画出来。合上书半年后,瞥一眼这张卡,那幅画面回来——这是「没白读」的物证。 > 图解 block 用手绘解释的风格、白底黑墨,**精确呈现意向画面**;继刚墨像(`assets/ljg-portrait.png`,由其头像抠底而成)是可选构图件,仅当画面里本来有个「你」才嵌入。完整设计历程见 `~/.claude/PAI/MEMORY/WORK/ljg-oneliner-design/ISA.md`。 Developers invoke ljg-library during build/integrations work for ai & agent building tasks. The skill documents triggers, prerequisites, and step-by-step workflows grounded in SKILL.md. Compatible with Claude Code, Cursor, and Codex agent runtimes that load marketplace skills.1.8kinstalls

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.