literature-search
lingzhi227/agent-research-skills
Search academic literature across Semantic Scholar, arXiv, and OpenAlex with structured results and BibTeX export.
What is literature-search?
Queries multiple academic databases to find relevant papers with metadata including title, authors, year, venue, abstract, citations, and BibTeX. Use when you need to discover papers, verify related work, or compile a bibliography for research.
- Search Semantic Scholar (best for ML/AI with BibTeX support)
- Query arXiv for latest preprints (< 3 months old)
- Access OpenAlex for broadest coverage without API key
- Merge and deduplicate results across sources
- Rank papers by citations, recency, venue quality, and relevance
- Generate BibTeX references from results
How to install literature-search
npx skills add https://github.com/lingzhi227/agent-research-skills --skill literature-search- Semantic Scholar API key (optional but recommended for best results)
- Python environment with requests library
- Access to internet for API calls
How to use literature-search
- 1.Provide a natural language search query
- 2.The skill expands your query into 2-4 complementary searches
- 3.Results are fetched from Semantic Scholar (primary), arXiv, and optionally OpenAlex
- 4.Results are merged, deduplicated, and ranked by citations, recency, venue quality, and relevance
- 5.Review the structured results table with BibTeX keys
- 6.Export results to .bib file or download arXiv source files as needed
Use cases
- Finding papers on a specific research topic to understand the landscape
- Building a related work section by searching complementary queries
- Discovering recent preprints on emerging techniques
- Compiling a bibliography with automatic BibTeX generation
- Assessing paper quality by venue tier and citation count
- Researchers conducting literature reviews
- PhD students building background sections
- Engineers evaluating state-of-the-art methods
- Anyone writing papers requiring academic citations
literature-search FAQ
Semantic Scholar is best for ML/AI papers with BibTeX support. arXiv has the latest preprints. OpenAlex provides broadest coverage and requires no API key. The skill queries all three and merges results.
Semantic Scholar API key is optional but recommended for better results. OpenAlex and arXiv are free and require no authentication.
Yes. Use flags like --top-conferences, --venue NeurIPS ICML, --year-range 2022-2026, --peer-reviewed-only, and --min-citations N.
Papers are ranked by: citations (30%), recency (30%), venue quality (20%), and relevance (20%). Venues are tiered from Tier 1 (NeurIPS, ICML, ICLR) down to preprints.
Yes. Results are returned in JSONL format with BibTeX keys, and you can generate a .bib file using the bibtex_manager script.
Full instructions (SKILL.md)
Source of truth, from lingzhi227/agent-research-skills.
name: literature-search description: Search academic literature using Semantic Scholar, arXiv, and OpenAlex APIs. Returns structured JSONL with title, authors, year, venue, abstract, citations, and BibTeX. Use when the user needs to find papers, check related work, or build a bibliography. argument-hint: [search-query]
Literature Search
Search multiple academic databases to find relevant papers.
Input
$ARGUMENTS— The search query (natural language)
Scripts
Semantic Scholar (primary — best for ML/AI, has BibTeX)
python ~/.claude/skills/deep-research/scripts/search_semantic_scholar.py \
--query "QUERY" --max-results 20 --year-range 2022-2026 \
--api-key "$(grep S2_API_Key /Users/lingzhi/Code/keys.md 2>/dev/null | cut -d: -f2 | tr -d ' ')" \
-o results_s2.jsonl
Key flags: --peer-reviewed-only, --top-conferences, --min-citations N, --venue NeurIPS ICML
arXiv (latest preprints)
python ~/.claude/skills/deep-research/scripts/search_arxiv.py \
--query "QUERY" --max-results 10 -o results_arxiv.jsonl
OpenAlex (broadest coverage, free, no API key)
python ~/.claude/skills/literature-search/scripts/search_openalex.py \
--query "QUERY" --max-results 20 --year-range 2022-2026 \
--min-citations 5 -o results_openalex.jsonl
Merge & Deduplicate
python ~/.claude/skills/deep-research/scripts/paper_db.py merge \
--inputs results_s2.jsonl results_arxiv.jsonl results_openalex.jsonl \
--output merged.jsonl
CrossRef (DOI-based lookup, broadest type coverage)
python ~/.claude/skills/literature-search/scripts/search_crossref.py \
--query "QUERY" --rows 10 --output results_crossref.jsonl
Key flags: --bibtex (output .bib format), --rows N
Download arXiv Source (get .tex files)
python ~/.claude/skills/literature-search/scripts/download_arxiv_source.py \
--title "Paper Title" --output-dir arxiv_papers/
Key flags: --arxiv-id 1706.03762, --metadata, --max-results N
Generate BibTeX from results
python ~/.claude/skills/deep-research/scripts/bibtex_manager.py \
--jsonl merged.jsonl --output references.bib
Workflow
- Expand the user's query into 2-4 complementary search queries
- Run Semantic Scholar search (primary) with expanded queries
- Run arXiv for very recent preprints (< 3 months)
- Optionally run OpenAlex for broader coverage
- Merge and deduplicate results
- Rank by: citations (0.3) + recency (0.3) + venue quality (0.2) + relevance (0.2)
- Present structured results table
Venue Quality Tiers
Tier 1: NeurIPS, ICML, ICLR, ACL, EMNLP, NAACL, CVPR, ICCV, ECCV, KDD, AAAI, IJCAI, SIGIR, WWW
Tier 2: AISTATS, UAI, COLT, COLING, EACL, WACV, JMLR, TACL
Tier 3: Workshops, arXiv preprints — mark with (preprint)
Output Format
Present results as a table + detailed entries with BibTeX keys. Always note preprint status.
Related Skills
- Downstream: citation-management, literature-review, related-work-writing
- See also: deep-research, novelty-assessment
Related skills
More from lingzhi227/agent-research-skills and the wider catalog.
math-reasoning
Formal mathematical reasoning for research papers — derive equations, write proofs, formalize problem settings, select statistical tests, and generate LaTeX math notation. Use when the user needs mathematical derivations, theorem proofs, notation tables, or statistical analysis formalization.
novelty-assessment
Assess research idea novelty through systematic literature search. Multi-round search-evaluate loops with harsh critic persona. Binary novel/not-novel decision with justification. Use before committing to a research direction.
paper-assembly
Orchestrate the full paper pipeline end-to-end. Manage state propagation between phases (literature → plan → code → experiments → figures → tables → writing → review), support checkpointing and resumption. Use for assembling a complete paper from components.
paper-compilation
Compile LaTeX papers to PDF with automatic error detection, chktex style checking, and citation/reference validation. Runs the full pdflatex + bibtex pipeline. Use when the user wants to compile a paper, fix compilation errors, or debug LaTeX.
paper-revision
Revise papers based on reviewer feedback. Map reviewer concerns to specific sections, apply targeted edits, run additional experiments if needed, and verify improvements. Use after receiving peer review with revision requests.
paper-to-code
Convert an ML research paper into a complete, runnable code repository. 3-stage pipeline from Paper2Code — Planning (UML + dependency graph) → Analysis (per-file logic) → Coding (dependency-ordered generation). Use for reproducing paper methods.