Awesome Paper Research Skills

Awesome Paper Research Skills

English | 中文

A curated map of open-source paper-related AI skills and workflows: idea discovery, literature search, experiment-to-paper bridges, academic writing, citation verification, LaTeX/DOCX formatting, and peer review.

This repo is built for researchers, students, and AI-agent builders who want to study the best public SKILL.md packs and assemble a serious paper pipeline without opening fifty tabs.

Current Version

Current workflow version: v3.2.6.

• Paper Research Workflow v3.2.6
• Root Claude/Codex Skill
• How To Use Individual Modules
• Paper Publication Orchestrator Skill
• Machine-Readable Publication Pipeline
• Example Pipeline Run Record
• Personal Use Config
• Skill Package Architecture
• Frontier Update - 2026-06-04
• Skill Roadmap
• Installable Skill Index
• Paper Language And Logic Style Guide
• Top Journal Section Writing Playbook
• Top Journal Section Writing Playbook - Chinese
• Scientific Palettes
• Palette Preview SVG
• Paper Frontier Radar Skill
• Paper Five Reviewer Panel Skill
• Paper Research Workflow v3.0.0
• Nature Skills Deep Dive
• Public Repository Landscape - 2026-05-29
• How To Choose A Paper Workflow Stack
• Paper Research Workflow v0.2.0
• Frontier Review - 2026-05-17
• Changelog

Why This Exists

Paper-related AI skills are fragmenting fast. Some repositories are full skill libraries, some are single SKILL.md files, some are research automation pipelines, and some are citation or formatting utilities. This list organizes them by workflow stage so you can study, compare, fork, and compose practical paper skills.

The v3.2 direction follows the current Agent Skills ecosystem: keep skills portable, source-grounded, inspectable, and ready for gh skill-style installation as the GitHub CLI preview matures. Third-party skills should be treated like executable dependencies: record provenance, inspect instructions, and flag prompt-injection or data-exfiltration risks before adoption.

The language layer also includes an AI-voice cleanup pass: reduce decorative quotation marks, em dashes, parentheses, adverb overuse, generic transitions, and inflated novelty language before reviewer simulation.

The repository now has a machine-readable pipeline layer in pipeline/, example run records and artifacts in examples/, per-skill manifests and OpenAI metadata under skills/*/, and validation in .github/workflows/validate.yml, so the workflow is not only prose: scripts and agents can inspect expected stages, gates, artifacts, skill calls, and submission blockers.

Quick Map

flowchart LR
  A["Frontier radar"] --> B["Idea discovery"]
  B --> C["Literature search"]
  C --> D["Experiment planning"]
  D --> E["Experiment execution"]
  E --> F["Analysis and figures"]
  F --> G["Paper drafting"]
  G --> H["Language and logic audit"]
  H --> I["Citation and provenance verification"]
  I --> J["Five-reviewer agent panel"]
  J --> K["Revision and re-review"]
  K --> L["LaTeX / DOCX export"]
  L --> M["Submission QA"]

Top Skill Packs

Star and fork counts are observed signals from GitHub/search snippets at the time of curation. They move quickly, so treat them as ranking hints rather than live counters.

Pipeline References

These are not all skill packs, but they are important references when designing paper skills:

Maintained Here / 本仓库维护项目

Top Repositories By Publication Stage / 按论文发表流程分类 Top 仓库

Each category shows the top repositories sorted by stars first and forks second. 每个分类优先按 stars、其次按 forks 排序，首页展示 Top 20；完整 400 个仓库见 Publication Flow Repository Map。

Idea Discovery And Research Question / 选题发现与研究问题

Find topics, generate hypotheses, check novelty, and turn broad interests into researchable questions.
用于发现研究方向、生成假设、初步判断创新性，并把宽泛想法转化为可执行的研究问题。

Literature Search And Reading / 文献检索与论文阅读

Search papers, build reading lists, summarize PDFs, and organize literature review inputs.
用于检索论文、整理阅读列表、总结 PDF，并为 related work 和综述搭建资料库。

Citation Management And Source Verification / 引用管理与来源验证

Manage BibTeX, DOI metadata, citation graphs, references, and hallucination checks.
用于管理 BibTeX、DOI、参考文献元数据，检查引用错误、来源缺失和伪造引用风险。

Experiment Execution And Reproducibility / 实验执行与可复现性

Run experiments, track results, manage datasets, and keep work reproducible.
用于运行实验、记录结果、管理数据和模型版本，并保持论文实验可复现。

Analysis, Statistics, Figures And Tables / 数据分析、统计、图表与表格

Analyze data, create publication-quality figures, tables, schematics, and statistical reports.
用于完成统计分析、可视化、论文级图表、表格和实验结果报告。

…