academic-research
Search and fetch academic papers and archives via inference.sh CLI. Apps: arXiv Search, arXiv Paper, bioRxiv Search, bioRxiv Paper, ChemRxiv Search, ChemRxiv Paper, Internet Archive Search, Metadata, Download, Wayback Machine, You.com Search, Deep Research, Contents, Finance Research. Capabilities: preprint search with field prefixes and category filters, paper metadata and PDF download, medRxiv lookup, archive.org items, Wayback snapshots, cited deep research. Use for: literature review, paper discovery, citation lookup, research agents, summarizing papers, historical web pages. Triggers: arx
Try your first task
A source-grounded starter task has not been verified for this skill yet.
SKILL.md
name: academic-researchdescription: "Search and fetch academic papers and archives via inference.sh CLI. Apps: arXiv Search, arXiv Paper, bioRxiv Search, bioRxiv Paper, ChemRxiv Search, ChemRxiv Paper, Internet Archive Search, Metadata, Download, Wayback Machine, You.com Search, Deep Research, Contents, Finance Research. Capabilities: preprint search with field prefixes and category filters, paper metadata and PDF download, medRxiv lookup, archive.org items, Wayback snapshots, cited deep research. Use for: literature review, paper discovery, citation lookup, research agents, summarizing papers, historical web pages. Triggers: arxiv, arxiv search, biorxiv, medrxiv, chemrxiv, preprint, research paper, academic search, literature search, paper summary, internet archive, archive.org, wayback machine, deep research, you.com, scholarly search, scientific papers"allowed-tools: Bash(belt *)Install the belt CLI skill:
npx skills add belt-sh/cli
Academic Research
Search preprint servers, the Internet Archive and the web via inference.sh CLI.
Quick Start
Requires inference.sh CLI (
belt). Install instructions
belt login
# Search arXiv
belt app run arxiv/search --input '{"query": "ti:diffusion AND cat:cs.CV", "max_results": 5}'
Available Apps
Preprint Servers
| App | App ID | Description |
|---|---|---|
| arXiv Search | arxiv/search | Query with field prefixes, boolean operators, category filter, sorting |
| arXiv Paper | arxiv/paper | Full metadata for one arXiv ID, optional PDF download |
| bioRxiv Search | biorxiv/search | bioRxiv and medRxiv preprints by date range, optional category |
| bioRxiv Paper | biorxiv/paper | One bioRxiv or medRxiv paper by DOI, optional PDF download |
| ChemRxiv Search | chemrxiv/search | ChemRxiv preprints by keyword (via Crossref) |
| ChemRxiv Paper | chemrxiv/paper | One ChemRxiv paper by DOI (via Crossref) |
Internet Archive
| App | App ID | Description |
|---|---|---|
| Archive Search | archive/search | Search archive.org with Lucene-like syntax |
| Archive Metadata | archive/metadata | Item metadata and file list |
| Archive Download | archive/download | Download one file from an item |
| Wayback | archive/wayback | Check whether a URL has a Wayback Machine snapshot |
You.com
| App | App ID | Description |
|---|---|---|
| Web Search | you/search | Web search with snippets and domain filters |
| Deep Research | you/research | Multi-step web research with cited sources |
| Contents | you/contents | Clean markdown or HTML from up to 10 URLs |
| Finance Research | you/finance-research | Financial research with filings and macro data |
Examples
arXiv Search
query supports field prefixes ti:, au:, abs:, cat:, all: and the operators AND, OR, ANDNOT.
belt app run arxiv/search --input '{
"query": "abs:\"retrieval augmented generation\" ANDNOT ti:survey",
"category": "cs.CL",
"sort_by": "submittedDate",
"sort_order": "descending",
"max_results": 20
}'
sort_by: relevance, lastUpdatedDate, submittedDate. Paginate with start.
arXiv Paper (with PDF)
belt app run arxiv/paper --input '{"arxiv_id": "2301.07041", "fetch_pdf": true}'
Returns paper metadata (title, authors, abstract, links) and pdf when requested.
bioRxiv / medRxiv Search
Search is by date range, not keyword. Filter with category.
belt app run biorxiv/search --input '{
"start_date": "2026-10-01",
"end_date": "2026-10-31",
"category": "neuroscience",
"server": "biorxiv"
}'
Set "server": "medrxiv" for medRxiv. Paginate with the returned cursor.
bioRxiv Paper
belt app run biorxiv/paper --input '{"doi": "10.1101/2023.01.01.123456", "fetch_pdf": true}'
ChemRxiv
belt app run chemrxiv/search --input '{"query": "metal organic frameworks CO2 capture", "rows": 10, "sort": "published-desc"}'
belt app run chemrxiv/paper --input '{"doi": "10.26434/chemrxiv-2023-fw8n4"}'
sort: relevance, published-desc, published-asc.
Internet Archive
# Search items (mediatype: texts, movies, audio, software, image, ...)
belt app run archive/search --input '{"query": "darwin origin of species", "mediatype": "texts", "sort": "downloads desc", "rows": 5}'
# Item metadata and file list
belt app run archive/metadata --input '{"identifier": "greatgatsby00fitzgerald"}'
# Download one file from the item (use a filename from the metadata file list)
belt app run archive/download --input '{"identifier": "<identifier>", "filename": "<filename>"}'
# Wayback snapshot (timestamp format YYYYMMDDhhmmss, optional)
belt app run archive/wayback --input '{"url": "example.com", "timestamp": "20150101000000"}'
You.com Search and Contents
belt app run you/search --input '{
"query": "protein structure prediction benchmarks",
"count": 10,
"include_domains": ["nature.com", "science.org"],
"recent_past_month": true
}'
belt app run you/contents --input '{"urls": ["https://arxiv.org/abs/2301.07041"], "format": "markdown"}'
You.com Deep Research
belt app run you/research --input '{
"query": "What are the main approaches to long-context evaluation in LLMs since 2025?",
"research_effort": "deep"
}'
research_effort: lite, standard (default), deep, exhaustive. Returns answer and sources.
You.com Finance Research
belt app run you/finance-research --input '{
"query": "Compare R&D spending trends of the top three semiconductor companies over the last 5 years",
"research_effort": "deep"
}'
research_effort: deep (default), exhaustive.
Workflow: Search Papers + Summarize
# 1. Find recent papers
belt app run arxiv/search --input '{
"query": "abs:\"speculative decoding\"",
"sort_by": "submittedDate",
"max_results": 10
}' --save papers.json
# 2. Pull full metadata for the most relevant one
belt app run arxiv/paper --input '{"arxiv_id": "<arxiv-id-from-results>"}' --save paper.json
# 3. Summarize with Claude Sonnet
belt app run anthropic/claude-sonnet-5-5 --input '{
"system_prompt": "You are a research assistant. Cite arXiv IDs for every claim.",
"text": "Summarize the main methods, results and open problems across these papers:\n\n<papers.json results>"
}'
For PDFs, pass the downloaded file to the LLM's files input instead of pasting text.
Pricing
| App | Pricing |
|---|---|
you/search | $0.005/call |
you/research | $0.001-$0.15/call (pass-through) |
you/finance-research | $0.11/request |
you/contents | Pass-through at upstream cost |
Check any app with belt app pricing <app-id>.
Related Skills
# Full platform skill (all apps)
npx skills add inference-sh/skills@infsh-cli
# Web search (Tavily, Exa)
npx skills add inference-sh/skills@web-search
# RAG pipelines (search + LLM)
npx skills add inference-sh/skills@ai-rag-pipeline
# Claude for summarization
npx skills add inference-sh/skills@claude
# All LLMs
npx skills add inference-sh/skills@llm-models
Browse all apps: belt app list
Documentation
- Adding Tools to Agents - Equip agents with search
- Building a Research Agent - LLM + search integration guide