name: deep-researcher
description: "Generate comprehensive 30-40 page academic research papers with full citations. Trigger: deep research, generate research paper, academic paper, literature review, research report, 30-40 page paper, comprehensive analysis, full citations, scholarly work."
metadata:
builtinskillversion: "1.1"
openclaw_native: true
replaces: knowledge-digest
triggers:

• "deep research"
• "generate research paper"
• "academic paper"
• "literature review"
• "research report"
• "comprehensive analysis"
• "scholarly paper"
• "long research paper"
• "research paper 30 pages"
• "full citations research"

Deep Researcher — Academic Research Paper Generator

Generate comprehensive, academic-grade research papers (30-40 pages) with 40-80 unique citations, following a rigorous 7-stage workflow. Adapts to any field — AI, medicine, economics, social sciences, engineering, and more.

Workflow Overview

STAGE 1: Topic Analysis      → Decompose topic into sub-questions
STAGE 2: Source Discovery    → Query academic & industry databases
STAGE 3: Content Synthesis  → Extract, summarize, map source relationships
STAGE 4: Cross-Verification  → Triangulate claims, verify facts
STAGE 5: Content Expansion   → Fill gaps, add case studies, data
STAGE 6: Synthesis & Writing → Assemble paper chapter-by-chapter
STAGE 7: Refinement & QA     → Polish, format citations, validate

Stage 1: Topic Analysis

Decompose the research topic into 12-15 subtopics. Identify:

• Primary research questions
• Scope boundaries (time period, geography, industry)
• Feasibility for 30-40 page scope
• Key theories and seminal works to anchor the paper

Output: Topic Deconstruction Report with subtopics, research questions, and knowledge gaps.

Stage 2: Source Discovery

Query multiple source categories using OpenClaw's native tools:

| Category | Sources | Tool |
|----------|---------|------|
| Academic | arXiv, Google Scholar, PubMed, Semantic Scholar | batch_web_search |
| Economic | World Bank API, IMF, OECD Stats | batch_web_search |
| Industry | McKinsey Insights, Statista, Gartner | batch_web_search + extract_content_from_websites |
| Code/AI | GitHub, Hugging Face, arXiv (CS) | batch_web_search |
| News | Reuters, BBC, RSS feeds | batch_web_search |
| Patents | Google Patents | batch_web_search |

Tool: batch_web_search (up to 10 concurrent queries)

For each search, extract: title, authors, year, DOI/URL, abstract, key findings.

Output: 40-80 candidate sources organized by category and relevance.

Stage 3: Content Synthesis

For each major source:

• Extract core contribution, methodology, key findings (3-5 bullet points), limitations
• Group by theme, methodology, and chapter alignment
• Identify patterns: recurring themes, contradictions, gaps

Output: Synthesized source notes (150-200 words per source), cross-reference map, draft literature review.

Stage 4: Cross-Verification

• Every factual claim must be backed by ≥1 source
• Critical claims (statistics, dates, specific findings) verified against 2-3 independent sources
• Flag sources with potential bias or industry funding
• Identify underrepresented perspectives

Output: Verification log, triangulation matrix, bias assessment.

Stage 5: Content Expansion

• Search for case studies, historical precedents, comparative analyses
• Add quantitative data from World Bank, IMF, OECD where relevant
• Include expert viewpoints, industry reports, conference proceedings
• Aim for 15-20 distinct subtopics covered

Output: Expanded source list (+10-20 sources), comparative analysis, historical timeline.

Stage 6: Synthesis & Writing

Assemble the paper using the Standard Research Paper Structure below. Integrate citations in APA 7th format. Write 15,000-18,000 words targeting 30-40 pages.

Standard Research Paper Structure

1. Title Page          (clear title, keywords, date)
2. Abstract            (300-500 words, 3-5 keywords)
3. Executive Summary   (1-2 pages, key takeaways for decision-makers)
4. Chapter 1: Introduction        (3-4 pages)
   - Background & context
   - Problem statement
   - Research objectives & questions
   - Significance & scope
5. Chapter 2: Literature Review   (6-8 pages)
   - Theoretical framework
   - Key themes (organized thematically)
   - Major studies & seminal works
   - Gaps in existing research
6. Chapter 3: Methodology         (4-5 pages)
   - Research design (qualitative/quantitative/mixed)
   - Data sources & search strategy
   - Inclusion/exclusion criteria
   - Analysis techniques & AI tools used
7. Chapter 4: Data Collection     (3-4 pages)
   - Sample/data description
   - Collection procedures
   - Ethical considerations
8. Chapter 5: Analysis & Findings  (8-10 pages)
   - Descriptive findings
   - Quantitative/qualitative results
   - Comparative and longitudinal analysis
   - Visual elements (tables, figures)
9. Chapter 6: Discussion          (3-4 pages)
   - Interpretation of key findings
   - Theoretical & practical implications
   - Limitations & counterarguments
10. Chapter 7: Conclusion         (2-3 pages)
    - Summary of contributions
    - Actionable recommendations
    - Future research directions
11. References                    (5-8 pages, 40-80 sources)
12. Appendices                    (optional)

Citation style: APA 7th Edition (Author, Year) — default. Also supports MLA 9th and Chicago Notes/Bibliography.

Stage 7: Refinement & QA

Run the quality checklist:

Accuracy

• [ ] 0% hallucinated claims — every claim backed by ≥1 source
• [ ] All statistics cross-checked against primary sources
• [ ] All dates within ±1 day of source
• [ ] All DOIs resolve correctly

Completeness

• [ ] 30-40 page target (15,000-18,000 words)
• [ ] ≥40 unique citations
• [ ] 15-20 distinct subtopics covered
• [ ] All 7 chapters present
• [ ] 5+ figures/tables

Coverage

• [ ] 4+ different source types (academic, industry, news, government)
• [ ] 70%+ sources from last 5 years
• [ ] 3+ distinct viewpoints represented
• [ ] At least 1 counterargument documented

Citation Integrity

• [ ] All in-text citations appear in References
• [ ] All References have in-text citations
• [ ] No orphan URLs
• [ ] APA 7th formatting consistent throughout

Literary Quality

• [ ] Formal academic tone, no slang
• [ ] Third-person perspective
• [ ] Clear transitions between chapters
• [ ] No repetition or redundancy
• [ ] Consistent terminology

Output: Final polished paper + QA report

Output Formats

| Format | Description | Tool |
|--------|-------------|------|
| Markdown | Default, editable | Direct output |
| PDF | Academic submission | minimax-pdf skill |
| DOCX | Word processing | minimax-docx skill |

Request with: "[topic] — output as PDF" or "[topic] — output as DOCX"

APA 7th Citation Quick Reference

In-text: (Author, Year) or Author (Year) showed that...

Reference entry (Journal):

Author, A. A., & Author, B. B. (Year). Title of article. Journal Name, Volume(Issue), Page.Range. https://doi.org/xxxxx

Reference entry (Book):

Author, A. A. (Year). Title of book (Edition ed.). Publisher.

Reference entry (Web):

Author, A. A. (Year, Month Day). Title of page. Website Name. https://url

Citation Density Rule

Target: ≥1 citation per 150-200 words across the full paper. This ensures every claim is evidence-backed and academically rigorous.

Source Priority Matrix

| Priority | Source Types | Cost |
|----------|-------------|------|
| HIGH | arXiv, PubMed, World Bank, IMF, OECD, Hugging Face, GitHub | Free |
| MEDIUM | Google Scholar, IEEE, McKinsey, Gartner, Statista | Free/Premium |
| LOW | News feeds, Twitter, Reddit, news blogs | Free |

Always prioritize free, authoritative, open-access sources first.

Data Sources Registry

Academic

• arXiv: https://export.arxiv.org/api/query — cutting-edge AI/ML/theory, no API key
• Google Scholar: batch_web_search — comprehensive peer-reviewed coverage
• PubMed/PMC: https://eutils.ncbi.nlm.nih.gov/entrez/eutils/ — biomedical, life sciences
• Semantic Scholar: https://api.semanticscholar.org/ — CS academic sources

Economic & Policy

• World Bank: https://api.worldbank.org/v2/ — free, no key required
• IMF: https://api.imf.org/ — macroeconomic data
• OECD: https://stats.oecd.org/ — comparative policy data

Technology & AI

• Hugging Face: https://api.huggingface.co/ — ML models, datasets, papers
• GitHub API: https://api.github.com/ — code trends, repositories
• Google Patents: https://patents.google.com/ — innovation trends

Industry

• McKinsey Insights: Public articles and reports
• Gartner: Industry analysis (subscription)
• Statista: Statistics and market data (subscription)

Troubleshooting

| Issue | Solution |
|-------|----------|
| "Insufficient sources" | Expand keyword list; use Boolean operators (AND/OR/NOT); try alternative databases |
| "Page count too short" | Expand Stage 5 (content expansion); add case studies, comparative data, visual elements |
| "Citation gaps" | Return to Stage 2; search for missing angles; add industry reports and government data |
| "Hallucination risk" | Always verify facts via Cross-Verification stage; cite primary sources only |
| "Formatting inconsistent" | Run APA 7th reference check; ensure all in-text citations match reference list |

Iteration Triggers

• From Stage 4 → Stage 3: Cross-verification reveals hallucination or insufficient evidence
• From Stage 5 → Stage 2: Expansion search finds critical gaps in core coverage
• From Stage 6 → Stage 4: Writing reveals missing critical sources
• Page count <30: Expand Stage 5, then revisit Stage 6

Integration Notes

This skill replaces and significantly extends the knowledge-digest skill's research capabilities. Where knowledge-digest focuses on learning materials from existing documents, deep-researcher generates original academic research from primary and secondary sources across multiple databases.

Both skills can coexist — use deep-researcher for original paper creation, knowledge-digest for study aids from existing materials.