academic-research

An Agent Skills for terminal-based AI agents (Claude Code, OpenCode, Cursor, Codex, etc.) that provides a complete academic research workflow.

What it does

• Search papers across arXiv, Semantic Scholar, DBLP, and CORE — all free, no API key needed
• Download PDFs with GB/T 7714-inspired structured filenames (Author_Year_Title_Source-SourceID.pdf) and an idempotent papers/index.json index
• Parse PDFs deeply — metadata, sections, figures, equations, two-column layout
• Read interactively — summarize, translate, explain formulas in plain language
• Discuss research ideas with adaptive mentor/peer reviewer modes

Install

npx skills add cCelectc/academic-research

Or copy skills/academic-research/ into your agent's skills directory.

Quick Start

Once the skill is loaded by your agent, just talk to it:

> Find recent papers about agent memory mechanisms
> Download paper #3
> Parse it
> Explain the architecture in section 3
> What if we applied this to multimodal agents?

Structure

skills/academic-research/
├── SKILL.md                       # Main instructions
├── scripts/
│   ├── pyproject.toml             # Dependencies (requests, pypdf)
│   ├── search.py                  # Multi-source search CLI
│   ├── parse_pdf.py               # Deep PDF parser CLI
│   └── download.py                # Download, structured naming, index.json
└── references/
    ├── search-sources.md          # API details per source
    └── peer-review-guide.md       # Mentor/peer interaction modes

Requirements

• Python 3.12+
• uv recommended (auto-falls back to venv + pip)
• curl or wget for downloads
• Internet access for API calls

Scripts bootstrap their own virtual environment on first run — no manual setup.

Data Sources

Source	Coverage	Has Abstracts	Has PDF Links	Has Citations
arXiv	CS, Physics, Math, Bio	Yes	Yes (direct)	No
Semantic Scholar	200M+ papers, all disciplines	Yes	OA only	Yes
DBLP	CS bibliography	No	No	No
CORE	OA repositories worldwide	Partial	Yes (OA)	No

License

MIT