AgentSkillsCN

literature-search

搜索并检索科学文献。当您需要查找论文、研究某一主题、寻找引用文献、获取论文摘要,或进行文献综述时,可选用此技能。支持接入 Semantic Scholar、arXiv 以及其他学术数据库。

SKILL.md
--- frontmatter
name: literature-search
description: Search and retrieve scientific literature. Use when asked to find papers, research a topic, find citations, get paper abstracts, or conduct literature reviews. Accesses Semantic Scholar, arXiv, and other academic databases.
allowed-tools:
  - Read
  - Write
  - Bash
  - WebSearch
  - WebFetch

Scientific Literature Search

You are conducting scientific literature searches and reviews.

Available Resources

Semantic Scholar (Primary)

  • 225M+ papers indexed
  • Rich citation network
  • AI-powered relevance ranking
  • API access via MCP server

arXiv

  • Preprints in physics, materials science, chemistry
  • Open access
  • Latest research before peer review

PubMed

  • Biomedical and life sciences
  • Peer-reviewed publications

CrossRef

  • DOI resolution
  • Publication metadata

Search Strategies

Topic Search

When researching a topic:

  1. Start with broad search terms
  2. Refine based on initial results
  3. Follow citation networks (cited by, references)
  4. Look for review articles first

Specific Paper Search

When looking for a specific paper:

  1. Search by title (exact or partial)
  2. Search by author + keywords
  3. Search by DOI if known

Citation Analysis

  1. Find highly cited papers in the field
  2. Look at recent papers citing foundational work
  3. Identify key authors and groups

Using Semantic Scholar MCP

The Semantic Scholar MCP server provides:

  • Paper search
  • Author search
  • Citation information
  • Paper recommendations

Example queries:

code
Search for papers on "CO2 adsorption in MOFs"
Find recent papers by Author Name about topic
Get citations for paper with ID xxx

Using Web Search

For broader searches:

code
Search for "metal-organic framework CO2 capture" site:nature.com
Search for "LAMMPS force field" filetype:pdf

Manual API Access (Fallback)

Semantic Scholar API

bash
curl "https://api.semanticscholar.org/graph/v1/paper/search?query=machine+learning+materials&limit=10"

arXiv API

bash
curl "http://export.arxiv.org/api/query?search_query=all:materials+science&max_results=10"

Literature Review Workflow

  1. Define Scope

    • What specific question are you answering?
    • What time period? (last 5 years typical)
    • What subfields?
  2. Initial Search

    • Use 3-5 different keyword combinations
    • Note total results to gauge field size
  3. Screen Results

    • Read titles and abstracts
    • Flag relevant papers
    • Note key authors and journals
  4. Deep Dive

    • Read full text of key papers
    • Extract methods, parameters, findings
    • Build citation network
  5. Synthesize

    • Identify consensus and controversies
    • Note gaps in literature
    • Summarize for user

Extracting Information

From papers, extract:

  • Methods: Simulation software, parameters, conditions
  • Force fields: Which potentials used, parameters
  • Results: Key numerical values, trends
  • Limitations: What authors acknowledge

Citation Format

Use consistent format:

code
Author1, Author2, et al. "Title." Journal Volume, Pages (Year). DOI: xxx

Saving Results

Save literature search results to:

code
workspaces/project-name/literature/
├── search-results.md       # Summary of search
├── key-papers.md           # Annotated bibliography
├── extracted-parameters.md # Force field params, etc.
└── pdfs/                   # Downloaded papers (if permitted)

Web Download (via Playwright)

For downloading papers or supplementary info:

code
Use playwright to navigate to [URL] and download the PDF

Note: Respect copyright and access restrictions.

Best Practices

  1. Document Everything: Record search terms, dates, result counts
  2. Check Recency: Prefer recent papers unless seeking foundational work
  3. Verify Citations: Cross-check important claims
  4. Look for Reviews: Start with review articles for new topics
  5. Follow Authors: Key researchers often have related work
  6. Check Preprints: arXiv may have newer versions

Common Queries

Materials Science

  • Force field parameters for [material]
  • DFT study of [property] in [material]
  • Molecular dynamics of [process]

Computational Methods

  • Best practices for [simulation type]
  • Convergence testing for [property]
  • Comparison of [method A] vs [method B]

Specific Materials

  • [Material] synthesis and characterization
  • [Material] applications in [field]
  • [Material] properties: [specific property]