papers

---
name: "papers"
description: "Searches academic literature via arXiv, Semantic Scholar, and open-access PDFs. Use when building literature reviews or finding formal research on a topic."
license: "MIT"
---

# Academic Papers Search

## When To Use

- Finding academic papers, citations, or formal research
- Building literature reviews or citation chains

## When NOT To Use

- Community opinions (use `/tome:discourse`)
- Code implementations (use `/tome:code-search`)

Search arXiv, Semantic Scholar, and open-access sources.

## Sources (Priority Order)

1. arXiv API (free, unlimited metadata)
2. Semantic Scholar API (100 req/5min, citation graphs)
3. Unpaywall (legal free version discovery)
4. CORE.ac.uk (open-access aggregator)
5. PubMed Central (biomedical)
6. Author preprint pages (WebSearch fallback)

## PDF Processing

After acquiring a paper URL or local file path, convert
the PDF to markdown for better extraction quality.

### Conversion (prefer markitdown)

Apply the `leyline:document-conversion` protocol:

1. **Try markitdown first**: call the MCP `convert_to_markdown`
   tool with the PDF URL or `file://` path. This produces
   structured markdown preserving tables, equations, figures,
   and section hierarchy.

2. **Fall back to Read tool** if markitdown is unavailable:
   - Read with `pages: "1-20"` for the first chunk
   - Continue with `pages: "21-40"` for longer papers
   - Continue in 20-page increments as needed
   - Note: tables and figures will not extract as cleanly

### Extraction Targets

From the converted markdown, extract:

- Abstract and key findings
- Methodology and experimental setup
- Results tables and figures
- Citation information (authors, year, venue)
- Key equations or formal definitions

## Fallback Guidance

When a paper is paywalled and no open version exists:
- Public library JSTOR/ProQuest access via library card
- DeepDyve article rental
- Inter-library loan request
- Direct author email template

## Exit Criteria

- [ ] At least one academic source queried in priority order
      (arXiv first, then Semantic Scholar, then Unpaywall/CORE/PubMed)
- [ ] PDF content converted via `leyline:document-conversion`
      protocol: markitdown MCP attempted first, Read-tool fallback
      used if markitdown is unavailable
- [ ] Each paper result includes citation info (authors, year, venue)
      and an abstract or key findings summary
- [ ] Paywalled papers with no open version trigger the fallback
      guidance (library, DeepDyve, ILL, author contact) rather than
      returning an empty result
- [ ] If all sources return zero results, this is stated explicitly
      rather than generating fabricated paper entries