A comprehensive library of AI agent skills for academic research and writing, designed to be used by autonomous agents to generate high-quality research outputs.
This repository uses a skill-based architecture to teach AI agents how to perform complex research tasks. Instead of generic prompting, we extract concrete, reusable patterns from top-tier conference papers (NeurIPS, ICLR, CVPR) and encapsulate them into "Skills".
.
├── docs/ # Skill definitions
│ ├── 01-academic-writing/ # Writing publication-ready papers
│ ├── 02-rebuttal-writing/ # Responding to reviewer comments
│ └── 03-presentation-slides/ # Creating conference presentations
├── src/ # Core implementation python scripts
├── data/ # Knowledge base (extracted skills, metadata)
└── references/ # Templates and detailed guides
- Python 3.8+
- OpenAI API Key (for skill extraction and generation)
pip install -r requirements.txtCollect data from top papers and extract writing patterns:
python3 src/scraper.py
python3 src/downloader.py
python3 src/processor.py
python3 src/skills_analyzer.py --buildUse the extracted skills to write a new paper from an idea:
# Generate outline
python3 src/main.py --mode outline --idea "Your research idea"
# Generate draft
python3 src/main.py --mode draft --outline output/outline.json
# Refine draft
python3 src/main.py --mode refine --draft output/draft.md- Purpose: Transform research ideas into full paper drafts.
- Source: Learned from Best Paper and Oral awards.
- Key Features: Structural templates, rhetorical pattern matching, argument verification.
- Purpose: Generate professional responses to reviewer critiques.
- Key Features: Tone management, point-by-point addressing logic.
- Purpose: Convert papers into slide decks for conferences.
- Key Features: Key point extraction, visual hierarchy planning.
MIT License
