JasperPWang/lab-codex-skills

最近记录的来源活动: 2026年10月21日 14:56

SkillsMP 收录数据更新: 2026年11月12日 12:52

已收集 skills: 54

GitHub 星标: 1
GitHub Forks: 0

用 AI 助手安装

复制这段 Prompt，发给你正在使用的 AI 助手。

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/JasperPWang/lab-codex-skills.

仓库 README

仓库 README 来源：README.md

查看 README 源文件

展开完整 README

# Lab Agent Skills This directory is the canonical source for the **Human Centric Intelligence Lab** skill bundle. It keeps user-maintained skills and vendored/downloaded skills together so the same research workflows can be shared across Macs and used by **any agent platform** that loads `SKILL.md` (Codex, Cursor, Antigravity, Claude Code, and similar). Public GitHub mirror: [https://github.com/JasperPWang/lab-codex-skills](https://github.com/JasperPWang/lab-codex-skills) (The repository name is historical; the bundle is **not** Codex-only.) ## What This Bundle Supports ### Agent platforms Skills are ordinary directories with a `SKILL.md`. Install scripts create **symlinks** into common discovery roots; you can also point any agent at this folder manually: | Platform | Typical link target | Script | | --- | --- | --- | | Codex | `~/.codex/skills/` | `scripts/link-to-codex-skills.sh` | | Cursor (project) | `<vault>/.cursor/skills/` | `scripts/link-to-cursor-skills.sh` | | Antigravity / agent-local | `<vault>/.agents/skills/` | `scripts/link-to-antigravity-skills.sh` | | Other agents | Set their skill root to this bundle or copy/symlink `SKILL.md` trees | — | `bash install.sh` links Codex always; when run from WorldModelVault’s `.tools/skills/`, it also links Cursor and Antigravity. ### After every skill edit (required) 1. Edit only the canonical `.tools/skills/` (or this repo root when working from the GitHub clone). 2. Run `bash install.sh` so Codex / Cursor / Antigravity discovery roots stay in sync on this machine. 3. Export and **push** to [lab-codex-skills](https://github.com/JasperPWang/lab-codex-skills) the same day—do not leave lab-facing changes only in the vault. 4. On other Macs: let iCloud sync the vault **or** `git pull` the clone, then run `install.sh` again so symlinks refresh. 5. On GPU servers (`3090`, `4090`, `A800`, `A6000`): each keeps `~/lab-codex-skills` + `~/.codex/skills` links. From a Mac that can SSH: ```bash bash ".tools/skills/scripts/sync-to-gpu-servers.sh" # or a subset: bash ".tools/skills/scripts/sync-to-gpu-servers.sh" 4090 A800 ``` Hosts that can reach GitHub use `git pull` + `install.sh`. **3090** often cannot reach GitHub; the script falls back to **rsync** from this machine. There is also a Codex-era union syncer at `.tools/scripts/sync-codex-skills-union-fast.py` (default hosts `4090 A800 A6000`, excludes `3090` unless passed). ### Note / document backends Research **content** skills are Markdown-first and platform-neutral. [`research-doc-workflow`](research-doc-workflow/SKILL.md) routes the same structure to the destination you name (URL or explicit path). Thin adapters apply only format deltas: | Backend | Adapter skill | Notes | | --- | --- | --- | | **Feishu / Lark** | [`feishu-doc-workflow`](feishu-doc-workflow/SKILL.md) | Native wiki/docs via `lark-cli`; whiteboards via `lark-whiteboard` when needed | | **Notion** | [`notion-doc-workflow`](notion-doc-workflow/SKILL.md) | Native pages/databases, properties, equations, captions; re-fetch verification | | **Obsidian / Markdown** | [`obsidian-doc-workflow`](obsidian-doc-workflow/SKILL.md) | Vault paths, callouts, frontmatter, wikilinks, relative assets; pair with `llm-wiki-skill` for wiki ingestion | Resolve the backend from the user’s **link or explicit instruction**. Do not silently default to Feishu or redirect one backend to another. Preserve the platform of an existing document. ## Rules - Put every editable skill under `.tools/skills/<skill-name>/` (or the root of this repo when cloned standalone). - Treat `~/.codex/skills/<skill-name>`, `.cursor/skills/`, and `.agents/skills/` as **compatibility symlinks** only—edit the canonical copy here. - Do not edit copied skill directories in multiple places. - Keep system, bundled, plugin-cache, and runtime skills outside this repository. Vendored third-party suites may live here when provenance is recorded in `THIRD_PARTY_NOTICES.md`. - After changing a skill, run its validator if present, or `scripts/validate-all.sh`. Before publishing or sharing, review: - `SKILL_MANIFEST.md` - `THIRD_PARTY_NOTICES.md` - `.gitignore` ## Copy-Paste Setup Prompt Paste into Codex, Cursor, Antigravity, or another agent on a teammate’s machine: ```text 请帮我配置 Human Centric Intelligence Lab 的 agent skill bundle。目标： 1. 从 GitHub 克隆公开仓库：https://github.com/JasperPWang/lab-codex-skills 2. 放到本机稳定位置，例如 ~/lab-codex-skills；若目录已存在，先 fetch/pull 更新，不要删除本地修改。 3. 运行 install.sh：链到 ~/.codex/skills；若在 Cursor/Antigravity 工作区使用，按 README 再链到 .cursor/skills 或 .agents/skills。 4. 运行 scripts/validate-all.sh，确认 skill 通过校验。 5. 确认能看到 research-doc-workflow、paper-card-delivery、paper-deep-dive、feishu-doc-workflow、notion-doc-workflow、obsidian-doc-workflow、survey-builder、ai-research-workflow、chinese-technical-writing 等常用入口。 6. 检查三类笔记后端：Obsidian/Markdown 是否可读写；Notion connector 是否已连接；若要编辑飞书原生文档，再检查 ~/.local/bin/lark-cli doctor。缺少后端时只说明需要配置，不要猜测或写入任何 token / secret。 7. 用中文总结：安装目录、校验结果、常用 prompt，以及 Feishu / Notion / Obsidian 是否就绪。约束： - 不要把 access token、refresh token、app secret、cookie、私有 Feishu 链接写进仓库或聊天。 - 同名普通目录先备份，不要直接覆盖；symlink 可以安全重建。 - 命令失败时先诊断，不要假装安装成功。 ``` On a new machine: ```bash # Standalone clone: git clone https://github.com/JasperPWang/lab-codex-skills.git ~/lab-codex-skills cd ~/lab-codex-skills bash install.sh # From WorldModelVault on this machine: bash ".tools/skills/install.sh" ``` Standalone Cursor / Antigravity linking examples: ```bash CURSOR_SKILL_ROOT=/path/to/project/.cursor/skills bash scripts/link-to-cursor-skills.sh AGENT_SKILL_ROOT=/path/to/project/.agents/skills bash scripts/link-to-antigravity-skills.sh ``` Validate: ```bash bash scripts/validate-all.sh # or from the vault: bash ".tools/skills/scripts/validate-all.sh" ``` Export a clean GitHub-ready copy: ```bash bash scripts/export-github-bundle.sh /tmp/lab-codex-skills ``` ## Document Backends (detail) Common research-content skills stay Markdown-first. `research-doc-workflow` keeps headings, lists, tables, links, code, and LaTeX aligned across backends. Only callouts, captions, properties/frontmatter, hierarchy, editable diagrams, citation chrome, uploads, and verification use thin adapters: - **Obsidian**: `obsidian-doc-workflow` (+ `llm-wiki-skill` when ingesting the Research Wiki). - **Notion**: `notion-doc-workflow` (native properties, one-paragraph paper-card metadata with `<br>`, re-fetch). - **Feishu/Lark**: `feishu-doc-workflow` + authenticated `lark-cli`. ### Feishu Native Editing Prerequisite ```bash ~/.local/bin/lark-cli doctor ``` Useful checks: ```bash ~/.local/bin/lark-cli docs +fetch --help ~/.local/bin/lark-cli docs +update --help ~/.local/bin/lark-cli docs +media-insert --help ~/.local/bin/lark-cli wiki --help ``` Configuration is user-local under `~/.lark-cli/`. Do not commit secrets into this repository. ## Integration Policy Personal lab skills own final deliverables. Downloaded skills can provide methods and checklists, but must not override the user’s semantic content contract or the selected Feishu / Notion / Obsidian representation. - Deep dives may borrow `nature-reader` methods, then deliver the lab package via `paper-deep-dive`. - `paper-deep-dive` is the single canonical deep-dive delivery standard. - Literature reviews may borrow `deep-research` methods, then deliver via `survey-builder`. - Paper cards remain governed by `paper-card-delivery`. - Cross-platform routing is governed by `research-doc-workflow`; each backend’s native blocks stay in its `*-doc-workflow` adapter. ## Research Workflow Skill Map Use short daily prompts. The agent should select the skill stack and only ask when missing context would change the result. | Research step | Daily prompt | Main skills | | --- | --- | --- | | Cross-platform page editing | `帮我整理/规范这个 Notion、Obsidian 或飞书页面` | `research-doc-workflow`, matching `*-doc-workflow`, `chinese-technical-writing` | | Paper cards | `帮我补全这个页面/笔记的 paper card` | `paper-card-delivery`, `research-doc-workflow`, matching adapter | | Single-paper deep dive | `帮我 deep dive 一下这篇论文，写到指定平台` | `paper-deep-dive`, `paper-card-delivery`, `research-doc-workflow`, matching adapter | | Literature tree / survey | `帮我把这个方向整理成 literature tree` | `survey-builder`, `ai-research-workflow`, `deep-research` | | Broad field exploration | `帮我调研这个方向，给我研究路线图` | `deep-research`, `academic-research-suite`, `nature-academic-search` | | Concept learning | `帮我讲清楚这个概念/方法` | `feynman-learning` | | Local wiki ingestion | `把这份材料纳入我的本地研究 wiki` | `llm-wiki-skill`, `obsidian-doc-workflow`, `chinese-technical-writing` | | Manuscript planning/writing | `帮我搭论文框架/写引言/改摘要` | `nature-writing`, `paper-spine`, `academic-paper` | | Manuscript polishing | `帮我润色这段英文论文文字` | `nature-polishing` | | Citation verification | `帮我核验这些引用是否真的支撑论断` | `cite-verify`, `nature-citation` | | Figures and plots | `帮我把这张图做成论文级图表` | `nature-figure` | | Statistics sanity check | `帮我检查这些实验表格和 p 值是否一致` | `stats-sanity` | | Experiment report / log | `帮我把这次运行整理成实验报告` | `experiment-report-writing`, `research-dev-standards` | | Experiment visualization | `继续当前项目，接入规范的实验可视化网页` | `research-viewer`, `ai-research-workflow`, `research-dev-standards` | | Reproducibility / release | `帮我整理这个项目的复现包` | `repro-pack` | | Mock review / rebuttal | `帮我从审稿人角度审一下/回审稿意见` | `academic-paper-reviewer`, `nature-reviewer`, `nature-response` | | Grant / proposal | `帮我搭这个基金/博士课题 proposal` | `grant-writer` | | Daily review | `帮我做今天的科研经营复盘` | `daily-research-review` | | Weekly summary | `帮我写/规范化这周的周总结或 GPT Weekly Pick` | `weekly-research-summary`, `research-doc-workflow`, `notion-doc-workflow` | ## Downloaded Skills Worth Using | Skill | Best use | Example prompt | | --- | --- | --- | | `deep-research` | Fuzzy new direction: questions, sources, synthesis, risks | `帮我调研 …，给我问题树、关键论文和研究空白。` | | `academic-research-suite` | Heavier academic pipeline coordination | `帮我把这个方向整理成可写论文的 research plan。` | | `nature-academic-search` | Multi-source search and BibTeX | `帮我检索 … 核心论文并导出 BibTeX。` | | `cite-verify` | Claim–citation alignment | `帮我核验这些引用是否支持每一句 claim。` | | `nature-reader` | Standalone bilingual reading; use `paper-deep-dive` for the lab package | `帮我做这篇论文的中英文对照精读。` | | `nature-writing` / `nature-polishing` | Drafting and English polish | `帮我组织 introduction / 润色这段文字。` | | `nature-figure` | Paper-grade figures | `用 Python 画论文级 figure。` | | `stats-sanity` | Numeric consistency | `检查表与正文数字是否一致。` | | `repro-pack` | Release / handoff | `整理成可复现包。` | | `paper-spine` | End-to-end manuscript build | `基于这些材料搭完整论文草稿。` | | `academic-paper-reviewer` / `nature-reviewer` / `nature-response` | Review and rebuttal | `从审稿人角度审 draft / 整理逐点回复。` | | `grant-writer` | Proposals | `搭 proposal 的 aims 和 approach。` | | `nature-paper2ppt` | Occasional PPTX from a paper | `做成 15 分钟组会 PPT。` | ## Common Prompts ### Daily Short Prompts ```text 帮我整理这个页面/笔记的 paper card： <FEISHU_OR_NOTION_URL_OR_OBSIDIAN_PATH> ``` ```text 帮我 deep dive 一下这篇论文： <PDF_OR_ARXIV_OR_PROJECT_URL> 目标：<FEISHU_OR_NOTION_URL_OR_OBSIDIAN_PATH> ``` ```text 帮我规范化这个研究页面/笔记，不要破坏已有图片和层级： <FEISHU_OR_NOTION_URL_OR_OBSIDIAN_PATH> ``` ```text 帮我把这个方向整理成 literature tree： <RESEARCH_TOPIC_OR_PAGE_URL_OR_OBSIDIAN_PATH> ``` ```text 帮我规范化这周的 Week Summary / GPT Weekly Pick（修公式与标题层级）： <NOTION_WEEK_PAGE_URL> ``` ### Feishu / Lark ```text 读取这个飞书页面，修改前先 fetch，修改后再 fetch 验证。不要破坏已有图片、表格、画板或多图排版。 <FEISHU_OR_LARK_URL> ``` ### Notion ```text 读取这个 Notion 页面或数据库记录，修改前先 fetch，修改后重新 fetch 验证。保留 properties、relations、原生 callout、公式、图片和 caption。 <NOTION_URL> ``` ### Obsidian / Markdown ```text 读取这个 Obsidian 笔记及相邻约定，做局部修改并在写后检查 frontmatter、wikilinks、公式和相对资源路径。 <OBSIDIAN_PATH> ``` ### Paper Cards / Deep Dive ```text 按照规范补全这个页面里的 paper card。必须先查看官方论文 HTML/PDF；中文 bullet 使用中文优先表达。 <FEISHU_OR_NOTION_URL_OR_OBSIDIAN_PATH> ``` ```text 对这篇论文做完整 deep dive，必须满足 paper-deep-dive 交付标准：页面属性用中文标题；正文开头是普通段落形式的英文标题、英文作者与单位，以及最新 arXiv PDF、Project Page、Code 三条链接，英文标题下面不放中文标题；随后只放完整原文中译稿，不默认写精读稿；公式保留 LaTeX；参考文献与正文引用按 deep-dive 链接合同；目标平台用对应 *-doc-workflow 写入并回读验证。 <PDF_OR_ARXIV_OR_PROJECT_URL> 目标：<FEISHU_OR_NOTION_URL_OR_OBSIDIAN_PATH> ``` ### Literature / Review / Wiki ```text 为这个方向建立 literature tree、novelty tree 和 challenge-insight tree。主题：<RESEARCH_TOPIC> ``` ```text 帮我做今天的科研经营复盘。最后写入当前 review 系统（Notion、Obsidian 或飞书），并输出明日三项关键 Todo。 ``` ```text 把这份材料纳入我的本地研究 wiki。保留来源锚点，生成中文正文与术语，并跑 lint/graph 检查。 <LOCAL_FILE_OR_SOURCE_URL> ``` ## Compatibility Aliases Install scripts also create: - `feishu-cli` → `feishu-doc-workflow` - `llm-wiki` → `llm-wiki-skill` - `paper-to-slides` → `paper-to-slides-skill`

正在加载仓库 README…

这个仓库中的 skills

其他高等院校教师软件开发工程师技术写作员桌面出版专家

6 个职业分类 · 已分类 56%

skill

职业分类

描述

更新

research-dev-standards

未分类

Enforces minimum-viable experimental design, evidence-based reasoning, test-driven development, tmux-first long-running sessions, Docker-first reproducible experiment environments, small incremental changes, 3D reconstruction result rendering, and…

2026-10-21

nature-figure

未分类

Submission-grade Nature/high-impact journal figure workflow for Python or R. Use whenever the user asks to create, revise, audit, or polish manuscript figures, multi-panel scientific plots, figures4papers-style matplotlib plots, or journal-ready SVG/PDF/TIFF…

原文语言：英语

2026-10-21

Use when summarizing or revising AI总结 sections in Obsidian clipping notes, especially Bilibili/video/subtitle notes under WorldModelVault. Produces concise Chinese summaries with a small vertical centered Mermaid flowchart, moderate structure, and…

原文语言：英语

2026-07-05

其他高等院校教师

Add strict Nature/CNS citations to manuscript text by splitting long passages into citable segments, searching only accepted flagship and subjournal titles from Nature Portfolio, the AAAS Science family, and Cell Press, filtering by publication time range,…

原文语言：英语

2026-07-04

已展示 40 / 54 个已收集 Skill。