Why Literature Search Skills Matter
If you have ever asked an AI agent to "find recent papers on CRISPR delivery mechanisms," you have probably received a confident-sounding list of papers that do not exist. Hallucinated citations are one of the most persistent failure modes of language models.
Literature search skills solve this by giving your agent direct access to real databases — PubMed, arXiv, OpenAlex, bioRxiv, and more. Instead of fabricating references, the agent queries actual APIs and returns verifiable results with DOIs, PMIDs, and links.
We surveyed 7 major skill registries (K-Dense, FreedomIntelligence, ClawBio, ARIS, Imbad0202, Orchestra, VoltAgent) and found 30+ skills related to literature search and academic research. This guide breaks them down by use case so you can pick exactly what you need.
Category 1: Single-Database Search
These skills connect your agent to one specific database. They are lightweight, often require zero configuration, and are the best starting point for most researchers.
| Skill | Registry | Database | API Key? | Best For |
|---|---|---|---|---|
pubmed-database | K-Dense (15.6K stars) | PubMed (NCBI E-utilities) | No | Boolean/MeSH queries, BioPython integration |
pubmed-search | FreedomIntelligence (1.5K stars) | PubMed (BioPython Entrez) | No | Quick keyword search with abstracts |
pubmed-summariser | ClawBio (476 stars) | PubMed (NCBI Entrez) | No | Gene/disease research briefs |
arxiv-database | K-Dense | arXiv (Atom API) | No | Search by keyword, author, ID, date, category |
arxiv-search | FreedomIntelligence | arXiv (Valyu API) | Valyu Key | Semantic search across all arXiv |
/arxiv | ARIS (2.9K stars) | arXiv API | No | Download PDFs to local library |
biorxiv-database | K-Dense | bioRxiv | No | Life science preprints |
medrxiv-search | FreedomIntelligence | medRxiv (Valyu) | Valyu Key | Medical preprint semantic search |
openalex-database | K-Dense | OpenAlex (240M papers) | No (free) | Largest open scholarly database |
Which single-database skill should you pick?
- Biomedical researchers: Start with
pubmed-databasefrom K-Dense. It has the most complete feature set including MeSH term support. - CS/ML researchers: Use
arxiv-databasefrom K-Dense. No API key needed, covers all arXiv categories. - Any discipline:
openalex-databasefrom K-Dense covers 240 million papers across all fields, completely free. - Quick gene/disease briefs:
pubmed-summariserfrom ClawBio is purpose-built for questions like "what is the latest research on gene X?"
Installation example:
clawhub install pubmed-database
clawhub install arxiv-database
clawhub install openalex-databaseCategory 2: Multi-Source Aggregated Search
When a single database is not enough, these skills search multiple sources simultaneously and merge the results.
| Skill | Registry | Sources | API Key? | Highlights |
|---|---|---|---|---|
literature-search | FreedomIntelligence | PubMed + arXiv + bioRxiv + medRxiv | Valyu Key | Four-source semantic search with aggregation |
biomedical-search | FreedomIntelligence | Above + ClinicalTrials.gov + FDA | Valyu Key | Adds clinical and regulatory data |
medical-research-toolkit | FreedomIntelligence | 14+ databases (ChEMBL, OpenTargets, OMIM, UniProt, KEGG...) | MCP config | Most comprehensive biomedical search |
lit-synthesizer | ClawBio | PubMed (MeSH) + bioRxiv + medRxiv | No | Citation graphs + research gap analysis |
Recommendations
- Most comprehensive:
medical-research-toolkitsearches 14 databases but requires MCP configuration. - Best value:
lit-synthesizerfrom ClawBio needs no extra API keys and adds citation graph construction and gap analysis on top of search. - Quick multi-source:
literature-searchfrom FreedomIntelligence searches four sources simultaneously but requires a Valyu key.
Category 3: Deep Full-Text Analysis
These go beyond titles and abstracts. They parse full papers, extract structured data, and can produce systematic reviews.
| Skill | Registry | Highlights | API Key? |
|---|---|---|---|
bgpt-paper-search | K-Dense / FreedomIntelligence | Full-text indexing, 25+ structured fields (methods, results, sample size, quality score) | BGPT MCP |
literature-review | K-Dense / FreedomIntelligence | Systematic review with PRISMA-style flow, multi-source cross-validation, APA/Nature output | No (chains other skills) |
/deep-research | Imbad0202 (862 stars) | 13-agent collaboration, 7 modes (full research, quick brief, paper review, systematic review, fact-check, Socratic dialogue, PRISMA + meta-analysis) | No |
tooluniverse-literature-deep-research | FreedomIntelligence | Target disambiguation, evidence grading, completeness checklist, testable hypothesis generation | MCP config |
Recommendations
- Deepest analysis:
/deep-researchfrom Imbad0202 runs a 13-agent PRISMA pipeline with meta-analysis support. Be warned: a single run can consume 200K+ input tokens. - Most structured output:
bgpt-paper-searchextracts 25+ fields from full text — ideal for evidence synthesis. - Best balance:
literature-reviewoffers solid depth at reasonable cost with properly formatted output.
Category 4: Citation Management
Because nothing undermines a paper faster than a fabricated reference.
| Skill | Registry | Sources | Highlights |
|---|---|---|---|
citation-management | K-Dense / FreedomIntelligence | Google Scholar + PubMed + CrossRef + arXiv | DOI/PMID/arXiv ID to BibTeX, prevents hallucinated citations |
abstract-searcher | VoltAgent Registry | arXiv + Semantic Scholar + CrossRef | Auto-fills abstracts in .bib files |
Category 5: Novelty Checking
| Skill | Registry | Highlights |
|---|---|---|
/novelty-check | ARIS | Extracts 3-5 core claims, searches each independently, adversarial cross-model review |
This is particularly valuable before submitting a paper — let your agent verify that your claims are actually novel.
Category 6: End-to-End Research Pipelines
These skills treat literature search as one step in a larger autonomous research workflow.
| Skill | Registry | Pipeline |
|---|---|---|
autoresearch | Orchestra (5.3K stars) | Literature survey -> topic selection -> experiments -> paper writing, supports overnight runs |
/idea-discovery | ARIS | Literature -> ideas -> novelty check -> peer review -> iterative refinement |
/deep-research + /academic-pipeline | Imbad0202 | 10-stage academic pipeline: research -> write -> review -> revise -> finalize |
Community-Recommended Registries
Beyond the major registries, these community projects offer specialized skills worth knowing about.
AIPOCH Medical Research Skills (37 stars)
A collection of original medical research skills covering the full evidence-based medicine workflow: search strategy design, database selection, evidence level prioritization, critical appraisal, and gap identification.
bash <(curl -s https://raw.githubusercontent.com/aipoch/medical-research-skills/main/scientific-skills/scripts/openclaw-install.sh)OpenClaw Master Skills — LeoYeAI (1.96K stars)
A curated collection updated weekly with 339+ skills. Notable entries:
academic-deep-research— transparent research process, not a black-box API wrapperdeep-research-pro— multi-source deep research agent with cited reports
clawhub install academic-deep-researchAMiner Open Skill — CanXiangCC
Access AMiner's scholarly platform with 28 APIs covering scholars, papers, institutions, journals, and patents.
aminer-data-search— full-featured (requires AMINER_API_KEY)aminer-free-search— free tier for discovery and filtering
General-Purpose Registries
| Skill | Registry | Source | API Key? | Notes |
|---|---|---|---|---|
deep-research (Gemini) | Antigravity (25K stars) | Google Gemini Deep Research API | GEMINI_API_KEY | $2-5 per run, 2-10 minutes, async support |
exa-search | Antigravity | Exa API (semantic search) | Exa Key | Has a dedicated "research papers" category |
latex-paper-conversion | Antigravity | Local pdflatex/bibtex | No | Convert between Springer, MDPI, IEEE, Nature formats |
Editor's Picks
Best Starting Point (Zero Config)
openalex-database from K-Dense
240 million papers, completely free, no API key required, covers all disciplines. If you install one skill today, make it this one.
clawhub install openalex-databaseBest for Biomedical Research
lit-synthesizer from ClawBio
Searches PubMed with MeSH terms plus bioRxiv and medRxiv preprints. No extra API keys. The bonus citation graph and research gap analysis features make it stand out from pure search tools.
Best for Deep Research
/deep-research from Imbad0202
A 13-agent PRISMA pipeline supporting 7 different modes including meta-analysis. Nothing else comes close in depth. Budget 200K+ tokens per run.
Best End-to-End Pipeline
/idea-discovery from ARIS
Goes from literature search to idea generation to novelty checking to peer review — all autonomously. Supports overnight runs for hands-off research exploration.
Quick Comparison Table
| Use Case | Recommended Skill | API Key? | Token Cost | Depth |
|---|---|---|---|---|
| Quick PubMed search | pubmed-database | No | Low | Basic |
| CS/ML paper search | arxiv-database | No | Low | Basic |
| Any discipline | openalex-database | No | Low | Basic |
| Multi-source biomedical | lit-synthesizer | No | Medium | Medium |
| 14-database biomedical | medical-research-toolkit | MCP | Medium | High |
| Systematic review | /deep-research | No | Very High (200K+) | Very High |
| Full research pipeline | /idea-discovery | No | High | End-to-End |
| Citation management | citation-management | No | Low | Utility |
| Novelty checking | /novelty-check | No | Medium | Utility |
Things to Watch Out For
- API key requirements: Some skills (Valyu, BGPT MCP, OpenRouter) need third-party keys. Check before you commit to a workflow.
- Shared upstream sources: K-Dense and FreedomIntelligence share many skills from the same upstream — do not install both copies.
- Token costs:
/deep-researchfrom Imbad0202 can consume 200K+ tokens in a single run. Great for thorough reviews, expensive for quick lookups. - Quality variance: VoltAgent Registry skills vary in quality. Prefer curated versions from K-Dense, FreedomIntelligence, or ClawBio when available.
Getting Started
The fastest path from zero to working literature search:
# 1. Install the universal search skill
clawhub install openalex-database
# 2. Add PubMed if you do biomedical research
clawhub install pubmed-database
# 3. Add arXiv if you do CS/ML research
clawhub install arxiv-database
# 4. Add citation management to prevent hallucinated references
clawhub install citation-managementOnce installed, just ask your agent naturally: "Find recent papers on transformer architectures for protein folding" — and it will query real databases instead of making things up.
Links
- OpenClaw Ecosystem
- K-Dense Registry
- ClawBio
- ARIS Skills
- Claw4Science — explore more OpenClaw ecosystem projects for science