ljg-skills README

The project's README from github.com/lijigang/ljg-skills. As one markdown file: README.md.

ljg-skills

我的 Codex 自定义技能集。

安装

使用 skills CLI 安装到 Codex:

# 安装全部技能(全局,org-mode 格式)
bunx skills add lijigang/ljg-skills -g -a codex --skill '*' -y

# 安装全部技能(Markdown 格式,适用于 Obsidian / VSCode / Notion 等)
bunx skills add lijigang/ljg-skills#md -g -a codex --skill '*' -y

# 安装单个技能
bunx skills add lijigang/ljg-skills -g -a codex --skill ljg-card -y

# 安装单个技能(Markdown 格式)
bunx skills add lijigang/ljg-skills#md -g -a codex --skill ljg-card -y

# 安装多个指定技能
bunx skills add lijigang/ljg-skills -g -a codex --skill ljg-card --skill ljg-learn -y

# 查看仓库中有哪些技能
bunx skills add lijigang/ljg-skills -l

参数说明:

参数 作用
-a codex 只安装给 Codex
-g 全局安装到 ~/.agents/skills/(推荐);不加则安装到项目级 .agents/skills/
--skill <name> 指定安装某个技能,可重复使用
--skill '*' 安装仓库内全部技能
#md 从 md branch 安装 Markdown 格式版本(默认为 org-mode)
-y 跳过交互确认
-l 仅列出可用技能,不安装

ljg-card 依赖

ljg-card 依赖 Playwright 截图,安装后需额外执行:

cd ~/.agents/skills/ljg-card && bun install && bunx playwright install chromium

替代方式:git clone

仓库根目录不是技能目录,clone 后还要把 skills/ 同步到 Codex。下面两种格式二选一:

# org-mode 版本(master)
git clone --branch master --depth 1 https://github.com/lijigang/ljg-skills.git "$HOME/code/ljg-skills"
mkdir -p "$HOME/.agents/skills"
rsync -a "$HOME/code/ljg-skills/skills/" "$HOME/.agents/skills/"

# Markdown 版本(md)
git clone --branch md --depth 1 https://github.com/lijigang/ljg-skills.git "$HOME/code/ljg-skills-md"
mkdir -p "$HOME/.agents/skills"
rsync -a "$HOME/code/ljg-skills-md/skills/" "$HOME/.agents/skills/"

技能

技能 说明
ljg-blind 盲区扫描 — 读取指定日期的 AI 对话,找出结构性思维盲区,并用微信读书章节精准补上
ljg-card 内容铸卡 — 将文本铸成长图 -l、原文保真全文卡 -f、漫画 -c、白板 -w 四种 PNG;-f 不生成新图,只呈现原稿自带图片,其余模式按各自图像合同生成位图
ljg-classic 古文精读 — 将原文、逐字注解、句义、章节意旨图与全章解读排成一张可连续阅读的长 PNG
ljg-learn 概念解剖 — 从八个方向切开一个概念(历史、辩证、现象、语言、形式、存在、美感、元反思),压成一句顿悟
ljg-paper 论文阅读 — 面向无行业背景的读者,从具体角色与动作讲清研究问题、作者贡献、结果如何得来及证据边界
ljg-book 拆书 — 讲清全书的主要内容、各条主线如何相连,以及事实、事件与后果怎样逐步改变理解;保留必要的多线发展
ljg-qa 信息提问机 — 把文章/论文/书的核心观点抽成 Q-A 链,Q 切要害,A 四段(结论 / 形式化 / 步骤 / 边界)
ljg-plain 白话引擎 — 把任何内容改写到聪明的十二岁小孩也能懂
ljg-rank 降秩引擎 — 给一个领域,找出背后不可再少的独立生成器
ljg-constraint 约束引擎 — 给一个领域/专业/角色,找出框住它的那几条约束(硬/软/自设三层),揪出被当成硬约束的假墙、指出哪条能重新定义
ljg-is 理解引擎 — 把名词的「是什么」与动词的「怎样运作」接起来,再推出认知修正与可执行判断
ljg-think 追本之箭 — 给一个观点或现象,纵向深钻到不可再分的本质
ljg-word 单词精通 — 深度拆解一个英语单词的核心语义和顿悟时刻
ljg-writes 写作引擎 — 把一个观点写成可理解、可迁移、经得住反例的中文文章
ljg-teach 体验式教学 — 先让学习者作出选择,用公平的意外显化认知缺口,命名后再以迁移局面复测
ljg-invest 投资分析 — 核心判断项目是否是一台「秩序创造机器」
ljg-read 伴读 — 陪你读任何文本,英文三层翻译(信达雅)+ 结构标注 + 深度提问 + 跨领域旁逸
ljg-relationship 关系分析 — 五层结构诊断 + 精神分析,通过对话引导帮用户"看见"关系真实结构
ljg-roundtable 圆桌讨论 — 一个议题一场圆桌:真实人物逐轮交锋,每轮收一张 ASCII 结构图,散场全文存档
ljg-structure 母题结构风洞 — 从表层问题找到反复出现的母题,提炼可迁移结构,用 ASCII 图说明关系并设计最小可逆实验
ljg-present 演讲铸造器 — 默认高桥流(一页一关键词、奶白底墨字);-s 标语流(VACAT/BIG STUDIOS 风:黑红双色块、ultra-bold、完整断言句撑屏)
ljg-push 推送引擎 — 把本地 ~/.agents/skills/ljg-* 一键同步到 github repo(master + md 双分支)

输出格式

技能提供两种输出格式,通过不同 branch 安装,功能完全相同:

Branch 格式 适用场景
master(默认) Org-mode(.org) Emacs / Denote 用户
md Markdown(.md) Obsidian / VSCode / Notion 等 Markdown 生态用户

ljg-skills (lijigang/ljg-skills)

Searching Database…

What is this?This is the author's personal collection of custom skills for the Codex AI agent, covering tasks like reading, writing, concept breakdown, paper summarization, and even relationship analysis.
Is this a standalone app or a skill library for Codex?It's a skill library: each skill is a folder of instructions (Org-mode or Markdown) that you install into Codex's skills directory, not a program you run by itself.
How do I install these skills into Codex?Use the 'skills' CLI with '-a codex' to target Codex and '-g' to install globally into ~/.agents/skills/ (recommended); use '--skill' to pick specific skills, or '--skill ''' to install all of them, with '-y' to skip confirmation or '-l' to just list what's available.
Does any skill need extra setup beyond the CLI install?Yes — the ljg-card skill relies on Playwright for taking screenshots and needs an extra install step after the main CLI install.
Can I just git clone the repo instead of using the CLI?You can, but the repo root isn't a skill folder by itself — after cloning you still have to manually sync the skills/ directory into Codex's skills path.
What kinds of tasks do individual skills handle?They span a wide range: ljg-qa turns an article into a chain of sharp Q-and-A pairs, ljg-card renders text into long shareable images, ljg-roundtable simulates a multi-voice debate with per-round ASCII diagrams, and ljg-classic produces annotated long-form reading images for classical Chinese texts.
Is this mainly for Chinese-language content?Most skills are built around Chinese reading, writing, and analysis workflows (e.g. classical text annotation, WeChat Reading cross-references), though a few like ljg-word focus on dissecting English vocabulary.
Why so many narrow skills instead of one general assistant?Each skill is scoped to a single well-defined task with its own output contract (e.g. fixed Q-A structure, specific image layout), aiming for consistent, structured results rather than generic open-ended answers.
Creator
lijigang
Size
2.5 MB
Stars
7.5K
Updated
CommitOct 12
Age
7mo
Commits · 90 Days