paper-search

---

name: paper-search description: Search for academic papers using natural language queries via Asta's semantic search across 225M+ papers. Use when finding research papers, exploring literature, or discovering relevant academic work on any scientific topic. allowed-tools: Bash, Read, Grep, Glob, TodoWrite

<quick_start> To search for papers:

Copy/asta:paper-search "transformer architectures for computer vision"

The skill will use the mcp__asta__snippet_search tool to find relevant passages, then format results with titles, authors, venues, and relevance snippets. </quick_start>

<success_criteria>

• Returns ranked list of relevant papers with metadata
• Each result includes: title, authors, year, venue, and relevance snippet
• Results are ordered by semantic relevance to the query
• Paper IDs (DOI/ArXiv) are included for verification </success_criteria>

Data Coverage:

• 225M+ papers from all scientific disciplines
• 12M+ full-text papers (285M+ passages)
• Sources: Semantic Scholar, arXiv, PubMed, major conferences

Step 2: Execute Search Use mcp__asta__snippet_search with parameters:

• query: The search query
• limit: Number of results (default 10, max 100)
• venues: Comma-separated venue filter (optional)

Step 3: Format Results For each result, extract and display:

• Paper title (bold)
• Authors (first 3 + "et al." if more)
• Year and venue
• Relevance snippet (the matching passage)
• Paper link and citation count

Step 4: Offer Follow-up If user wants more details on a specific paper, use mcp__asta__get_paper to fetch full metadata.

<output_format> Present results in this structure:

Copy## Paper Search Results for: "{query}"

**Found {n} relevant papers:**

1. **{Title}** ({Year})
   *{Author1}, {Author2}, et al.* — {Venue}
   > "{Relevant snippet from the paper...}"
   DOI: {doi} | Citations: {count}

2. **{Title}** ({Year})
   ...

</output_format>

Found 5 relevant papers:

1. Attention Is All You Need (2017) Vaswani, Ashish et al. — NeurIPS

   "We propose a new simple network architecture, the Transformer, based solely on attention mechanisms..." arXiv:1706.03762 | Citations: 95,000+

2. BERT: Pre-training of Deep Bidirectional Transformers (2019) Devlin, Jacob et al. — NAACL

   "BERT uses a multi-head self-attention mechanism that allows the model to jointly attend to information..." arXiv:1810.04805 | Citations: 78,000+

<anti_patterns>

• Don't guess paper details - Always use MCP tools to retrieve actual data
• Don't return too many results - Default to 10, respect user limits
• Don't omit paper IDs - Always include DOI/ArXiv for verification
• Don't fabricate citations - Only report actual citation counts from the API </anti_patterns>