deep-research（深入研究）

Deep Literature Research

Conduct a thorough literature search on a topic with verified citations. CRITICAL: Never fabricate citations. Every claim must have a verifiable source.

When to Use

• Starting a new project (research core concepts)
• Exploring methodology options
• Understanding the current state of a field
• Finding prior work on a specific technique

Usage

/deep_research [topic]
/deep_research normalization methods for RNA-seq
/deep_research machine learning in proteomics

The 5-Step Literature Review Process

Follow this framework for rigorous, reproducible literature reviews:

1. DEFINE SCOPE    → What are we looking for? Set boundaries.
2. SEARCH          → Systematic, documented search across databases.
3. EVALUATE        → Assess source quality and relevance.
4. SYNTHESIZE      → Identify themes, patterns, and gaps.
5. DOCUMENT        → Write findings with proper citations.

Execution Steps

1. Understand the Research Need (Define Scope)

If topic is vague, ask clarifying questions:

• "What specific aspect of [topic] are you most interested in?"
• "Are you looking for methodological approaches, theoretical background, or applications?"
• "Any specific time range? (e.g., last 5 years, seminal works)"

2. Generate Search Strategy

Break the topic into searchable concepts:

## Search Strategy for: [Topic]

### Core Concepts
1. [Primary concept] - synonyms: [alternatives]
2. [Secondary concept] - synonyms: [alternatives]
3. [Methodological aspect]

### Search Queries
- "[concept1] AND [concept2]"
- "[method] in [application domain]"
- "review [topic]" (for overview papers)

### Key Databases
- PubMed (biomedical)
- arXiv (computational, preprints)
- Google Scholar (broad)
- Semantic Scholar (AI-enhanced)

3. Evaluate Sources

Apply the ACRAP criteria to each source:

Quick Assessment:

• ✅ High quality: Peer-reviewed, reputable journal, clear methodology
• ⚠️ Use with caution: Preprints, conference papers, older works
• ❌ Avoid: Non-peer-reviewed, predatory journals, unsupported claims

4. Conduct Research

CRITICAL RULES:

1. ✅ Only include claims with verifiable sources
2. ✅ Format citations properly (Author et al., Year)
3. ✅ Include DOI or URL when available
4. ❌ NEVER fabricate or guess citations
5. ❌ NEVER make up author names, years, or findings
6. ✅ If no source exists, explicitly state: "Gap identified: No literature found on [specific topic]"
7. ✅ If uncertain, say: "Limited sources found. Manual verification recommended."

5. Organize Findings

Structure the output as:

# Literature Review: [Topic]

## Executive Summary
[2-3 sentence overview of the field]

## Current State of the Field

### [Theme 1]
[Synthesized findings with citations]

Key findings:
- Finding 1 (Author et al., Year)
- Finding 2 (Author et al., Year)

### [Theme 2]
[Synthesized findings]

### Methodological Approaches
[What methods are commonly used]

### Gaps in the Literature
- Gap 1: [What's missing or unexplored]
- Gap 2: [Contradictions or debates]

## Seminal Works
[Important foundational papers that should be cited]

## Recent Advances (Last 2-3 years)
[Most current developments]

## Implications for This Project
[How this literature relates to the user's aims]

## References

[Full BibTeX entries for all citations]

6. Save Outputs

Save to .research/literature/:

• [topic-slug].md - The literature summary
• [topic-slug].bib - BibTeX citations

Example:

.research/literature/
├── rna-seq-normalization.md
└── rna-seq-normalization.bib

7. Prompt Next Steps

After completing:

Literature review saved to .research/literature/[topic].md

Suggested next steps:
A) Run /write_background to draft your background section
B) Run /deep_research on another topic: [suggest related topic]
C) Review the findings and refine your project aims

Would you like to explore any of these sources in more detail?

Citation Format Standards

In-Text Citations

• Single author: (Smith, 2023)
• Two authors: (Smith & Jones, 2023)
• Three+ authors: (Smith et al., 2023)
• Multiple citations: (Smith, 2023; Jones, 2022)

BibTeX Format

@article{smith2023keyword,
  author = {Smith, John and Jones, Jane and Williams, Bob},
  title = {Title of the Paper},
  journal = {Journal Name},
  year = {2023},
  volume = {10},
  number = {2},
  pages = {123-145},
  doi = {10.1234/example.doi}
}

Handling Uncertainty

When sources are limited:

⚠️ Limited sources found on [specific topic].
Available sources cover [related area] but not [specific aspect].
Recommend:
- Manual search in [specific database]
- Consultation with domain expert
- This may represent a gap in the field

When sources conflict:

📊 Conflicting findings in the literature:
- Position A: [claim] (Author1, Year)
- Position B: [opposing claim] (Author2, Year)
Current consensus appears to favor [position] based on [evidence].

When no sources found:

🔍 Gap identified: No peer-reviewed literature found on [topic].
This could mean:
- Novel research opportunity
- Need for different search terms
- Topic may be covered under different terminology
Suggested alternative searches: [alternatives]

Quality Checks

Before finalizing, verify:

• Every factual claim has a citation
• All citations are real and verifiable
• BibTeX entries are complete
• DOIs are included where available
• No made-up author names or publication details
• Gaps and limitations are explicitly stated

Related Skills

• write-background - Use literature to draft background section
• next - Get next suggested step