---
name: search_skill
description: Academic literature search skill for finding papers from arXiv, Semantic Scholar, and HuggingFace
version: 1.0.0
author: PaperAgent Team
---

# Academic Literature Search Skill

This skill enables you to search for academic papers and literature from multiple sources. Use this skill when you need to find relevant research papers for a topic.

## Available Tools

You have access to the following search tools (registered in the Toolkit):

### 1. `search_arxiv`
Search for papers on arXiv preprint server.

**Parameters:**
- `query` (str, required): The search query string
- `max_results` (int, optional): Maximum number of results (default: 10, max: 50)
- `sort_by` (str, optional): Sort by "relevance" or "date"

**Returns:** List of papers with title, authors, abstract, URL, year, and categories.

**Example:**
```
search_arxiv(query="large language models", max_results=10, sort_by="relevance")
```

### 2. `search_semantic_scholar`
Search for papers on Semantic Scholar with citation information.

**Parameters:**
- `query` (str, required): The search query string
- `max_results` (int, optional): Maximum number of results (default: 10)
- `year_from` (int, optional): Filter papers from this year
- `year_to` (int, optional): Filter papers to this year

**Returns:** List of papers with title, authors, abstract, URL, year, citation count, and venue.

**Example:**
```
search_semantic_scholar(query="transformer architecture", max_results=20, year_from=2020)
```

### 3. `search_huggingface_papers`
Search for papers on HuggingFace Papers.

**Parameters:**
- `query` (str, required): The search query string
- `max_results` (int, optional): Maximum number of results (default: 10)

**Returns:** List of papers with title, authors, abstract, and URL.

**Example:**
```
search_huggingface_papers(query="multimodal learning", max_results=15)
```

### 4. `academic_search`
Multi-source academic search that combines results from multiple sources.

**Parameters:**
- `query` (str, required): The search query string
- `sources` (list[str], optional): Sources to search (default: ["arxiv", "semantic_scholar"])
- `max_results_per_source` (int, optional): Max results per source (default: 5)

**Returns:** Combined and deduplicated list of papers from all sources.

**Example:**
```
academic_search(query="neural networks", sources=["arxiv", "semantic_scholar"], max_results_per_source=10)
```

## Search Strategies

### Topic-based Search
For exploring a research topic:
1. Start with broad query terms
2. Use `academic_search` to get papers from multiple sources
3. Identify key papers and refine search based on terminology found

### Author-based Search
For tracking specific researchers:
1. Include author name in query: `"Author Name" topic`
2. Use Semantic Scholar for better author disambiguation

### Citation-based Search
For finding related work:
1. Search for seed papers on the topic
2. Use citation tools to trace references and citing papers

### Time-based Search
For finding recent advances:
1. Use `year_from` parameter in Semantic Scholar
2. Use `sort_by="date"` in arXiv

## Best Practices

1. **Formulate precise queries**: Use specific technical terms
2. **Combine sources**: Different databases have different coverage
3. **Filter by year**: For up-to-date research, filter recent papers
4. **Check citation counts**: High citation papers are often foundational
5. **Read abstracts first**: Quickly assess relevance before deep reading
6. **Save search results**: Keep track of found papers for later reference

## Output Format

Search results are returned in markdown format with:
- Paper title
- Authors (first 3 + "et al." if more)
- Publication year
- Citation count (when available)
- Venue (when available)
- Abstract (truncated to 500 chars)
- URL/DOI link