agentleFS
PilgrimChPilgrimCh/research-note/SKILL.md

research-note

Skill0 starsChanged 3 months agoMITView on GitHub

生成中文深度研究笔记、LaTeX/PDF、StudyHub 论文卡片和研究日志。适用于论文精读、 多篇论文综合、专题笔记、暑研阅读、研究问题沉淀。默认面向 Codex + Windows + Obsidian StudyHub 工作流;长论文不能压缩成短摘要,必须产出教学型、可复用的深度笔记。

3 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions3 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 3

    1. daloopa/investing★ 488same name
    2. life-itself/skills★ 2same name
    3. ahnchiyoon87/claude-skills★ 0same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 1 is distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use xelatex and biber. For Chinese filenames, avoid biber encoding problems by compiling with an ASCII job name:

    line 249
  2. Very specific:

    Run pdfinfo and confirm page count.

    line 260
  3. Very specific:

    Avoid math, commas, square brackets, or unmatched punctuation inside optional tcolorbox titles such as \begin{comparebox}[...].

    line 272
  4. Very specific:

    If Poppler prints a non-fatal warning such as Syntax Error: Weird page contents but renders PNG pages successfully, treat it as a warning …

    line 274
  5. Very specific:

    Do not deliver if the log contains !

    line 276
  6. Distinctive:

    Use these tcolorbox environments when available in the template:

    line 280

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Research Note — Codex/Windows + StudyHub 深度研究笔记

0. Core Goal

输出不是论文摘要,而是可复用的研究材料:

  1. 中文深度 LaTeX/PDF 研究笔记。
  2. StudyHub Markdown 论文卡片与研究状态更新。
  3. 对用户当前研究主线的具体连接和后续问题清单。

默认风格参考深度专题笔记:先补基础概念,再进入方法、理论、实验、对比、批判和研究启发。长论文或多论文任务不得压缩到 5-6 页,除非用户明确要求 brief。

1. Trigger And Mode

  • Day N: 按学习计划生成当天研究日志。
  • 专题 <主题>: 从基础到前沿的专题型笔记。
  • papers <pdf...> 或用户给出论文 PDF: 多篇论文精读/综合。
  • 用户提到 Obsidian、StudyHub、研究主线、暑研、与我研究的关联时:必须使用 StudyHub 记忆层。

如果用户没有指定模式:

  • 单篇论文: 默认 专题/论文精读。
  • 多篇相关论文: 默认 多论文综合专题。
  • 论文总页数 > 40 或任一论文 > 20 页: 默认 深度模式。

2. Mandatory StudyHub Memory Layer

先解析 STUDYHUB_ROOT,不要假设所有用户使用同一个绝对路径。按以下优先级:

  1. 用户在当前请求中明确提供的 vault 路径。
  2. skill 根目录旁的可选 local-config.yaml 中的 studyhub_root。
  3. 环境变量 STUDYHUB_ROOT。
  4. 仍无法解析时,先向用户询问 vault 路径,再进行任何 StudyHub 写入。

local-config.yaml 是仅限本机的配置,不得提交到公共仓库。格式参考 local-config.example.yaml。

涉及用户研究方向、暑研、paper reading、“与我的研究关联”、后续计划时,先读:

  1. {STUDYHUB_ROOT}\Home\Codex Runtime Memory.md
  2. {STUDYHUB_ROOT}\Research\Sustech research\Sustech research.md
  3. 若存在,再读:
    • {STUDYHUB_ROOT}\Research\Sustech research\Current Research State.md
    • {STUDYHUB_ROOT}\Research\Sustech research\Paper Reading Index.md
    • {STUDYHUB_ROOT}\Research\Sustech research\Research Questions.md
    • {STUDYHUB_ROOT}\Research\Sustech research\Open Problems.md
    • {STUDYHUB_ROOT}\Research\Bandits\Bandits.md

不要把“与我的研究的关联”写成泛泛而谈。必须连接用户当前主线:

  • statistics / bandits / offline evaluation
  • LLM exploration-exploitation
  • in-context RL
  • LLM agents as decision-making systems
  • SUSTech summer research with Prof. Fang Kong

3. Output Contract

每次深度研究任务默认输出:

  1. deep_research_note.tex
  2. deep_research_note.pdf
  3. references.bib
  4. StudyHub paper card(s), Markdown:
    • {STUDYHUB_ROOT}\Research\Sustech research\Paper Notes\<short-title>.md
  5. StudyHub updates where relevant:
    • Paper Reading Index.md
    • Research Questions.md
    • Open Problems.md
    • Weekly Research Log.md

如果时间或环境不允许全部完成,优先级为:

PDF note > references.bib > paper cards > index/log updates.

3.1 Workspace Layout Contract

当用户给出项目目录或工作目录时,必须保持根目录干净。

项目根目录只允许出现:

  1. 用户明确提供的原始源文件,例如论文 PDF。
  2. 一个最终成品文件夹,例如 final_deliverables。
  3. 一个中间文件夹,例如 intermediate_files。

项目根目录禁止出现:

  • 抽取出的 .txt 文本。
  • 仅用于参考的样例 PDF 副本。
  • 页面渲染 PNG。
  • LaTeX 源码、.aux、.bbl、.bcf、.log、.out、.toc、.run.xml。
  • BibTeX 文件。
  • 临时脚本。
  • 旧版本输出文件夹。

所有非最终 research-note 产物必须进入中间文件夹,例如:

intermediate_files/
  sources_text/
  reference_samples/
  research_note/
  render_checks/

最终成品文件夹只放用户会直接打开或提交的文件。若与 slide skill 同时使用,两个 skill 必须共用同一个中间文件夹和同一个最终成品文件夹。

4. Depth Requirements

Brief Mode

仅在用户明确要求“简短/brief/快速总结”时使用。

  • 3-6 页 PDF。
  • 适合快速扫读。

Standard Mode

  • 单篇短论文或普通阅读: 8-12 页 PDF。
  • 至少覆盖 problem, method, theory/algorithm, experiments, critique, research connection.

Deep Mode

触发条件:任一论文 > 20 页,或多篇论文总页数 > 40,或用户要做 presentation / 暑研材料。

  • 目标 12-25 页 PDF。
  • 多篇论文时,每篇论文至少 4-6 页实质内容。
  • 必须逐节覆盖:
    1. 背景与前置概念
    2. 论文问题设定
    3. 方法机制
    4. 关键公式/算法/理论
    5. 实验设计与结果
    6. 与相关方法对比
    7. 批判性分析
    8. 与用户研究主线的连接
    9. 可执行后续问题

禁止把 40+ 页材料压缩成只含摘要、优缺点和几条启发的短文。

5. Recommended Structure

Single-Paper Deep Topic

\part{I} 基础概念
1. 问题背景
2. 关键前置知识

\part{II} 方法详解
3. 核心设计
4. 算法/模型/训练目标
5. 部署或推断流程

\part{III} 理论与实验
6. 理论结果或机制解释
7. 实验设置
8. 关键结果与现象

\part{IV} 深入理解与对比
9. 与相关方法对比
10. 失败模式与边界条件

\part{V} 批判性分析与研究展望
11. 批判性评价
12. 与我的研究的关联
13. 后续问题与计划
14. 自检清单
\printbibliography

Multi-Paper Synthesis

\part{I} 背景与统一问题
\part{II} Paper A 精读
\part{III} Paper B 精读
\part{IV} 横向综合:共同问题、关键差异、互补关系
\part{V} 对用户研究主线的启发
\part{VI} 批判性分析、后续实验、开放问题
\printbibliography

6. Content Rules

  • 全部中文撰写,术语可保留英文括注。
  • 数学公式必须用 LaTeX。
  • 每个新概念配一个直觉解释,优先使用 intuitionbox。
  • 关键公式需要说明每个符号的含义。
  • 关键理论不要只抄 theorem,要解释假设、结论、直觉和局限。
  • 实验部分必须写清:数据/环境、baseline、metric、主要发现、作者想证明什么。
  • 跨方法对比必须用 comparebox + booktabs 表格。
  • 批判性分析必须四段式:
    • 优势
    • 局限
    • 未解决问题
    • 研究机会
  • 与我研究的关联必须给出具体可执行连接,不写空泛“有启发”。

7. StudyHub Markdown Rules

StudyHub 笔记使用 Obsidian-renderable Markdown。

  • inline math: $...$
  • display math: $$...$$
  • 不使用 \(...\) 或 \[...\]。

Paper card 模板:

---
type: paper-note
status: read
topic: <topic>
created: <date>
source: <pdf/arxiv>
---

# <Paper Title>

## One-Sentence Contribution

## Problem Setting

## Method

## Theory / Algorithm

## Experiments

## Limitations

## Connection to My SUSTech Line

## Possible Research Questions

## Terms To Remember

8. Windows/Codex Compilation Protocol

Use xelatex and biber. For Chinese filenames, avoid biber encoding problems by compiling with an ASCII job name:

xelatex -jobname=deep_note -interaction=nonstopmode "<中文文件名>.tex"
biber deep_note
xelatex -jobname=deep_note -interaction=nonstopmode "<中文文件名>.tex"
xelatex -jobname=deep_note -interaction=nonstopmode "<中文文件名>.tex"

After compilation:

  1. Run pdfinfo and confirm page count.
  2. Render first 3 pages and last page with pdftoppm.
  3. Check logs for:
    • undefined citations
    • empty bibliography
    • fatal errors
    • severe overfull boxes

Warnings about minor fancyhdr headheight or small underfull boxes are acceptable if the PDF renders cleanly.

Known Windows/LaTeX pitfalls

  • Avoid math, commas, square brackets, or unmatched punctuation inside optional tcolorbox titles such as \begin{comparebox}[...]. Titles like [$(k,v)$ 设计轴] can be parsed as pgfkeys options and break compilation. Use plain titles such as [k-v 设计轴].
  • Compile Chinese .tex files with an ASCII -jobname; otherwise biber may look for mojibake .bcf filenames.
  • If Poppler prints a non-fatal warning such as Syntax Error: Weird page contents but renders PNG pages successfully, treat it as a warning and inspect the images.
  • Tables with long English identifiers (ScienceWorld, REINFORCE+LoRA, Neural-LinLogUCB) often cause underfull boxes. Prefer tabularx, shorter labels, or split into multiple tables.
  • Do not deliver if the log contains ! LaTeX Error, ! Package ... Error, undefined references, or Empty bibliography after the final run.

9. Box Usage

Use these tcolorbox environments when available in the template:

BoxPurpose
keybox核心公式/论文主张
intuitionbox新概念直觉
derivbox推导过程
comparebox方法对比表
flowbox流程图/研究路线
warnbox局限、陷阱、假设
critiquebox批判性分析
notebox补充说明

10. Final Response

报告:

  • PDF 路径与页数。
  • .tex 与 references.bib 路径。
  • StudyHub 更新了哪些文件。
  • 编译/渲染验证结果。
  • 如果因环境限制跳过了某项,明确说明。

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.