Discover academic literature through citation-network analysis instead of keyword matching alone. Inciteful builds a citation graph around one or more seed papers and ranks the network by similarity (co-citation + Adamic-Adar), PageRank, and recency. IDs are OpenAlex IDs, so results interop with the paper-search skill (use its paper.sh for abstracts, which Inciteful does not return).
Setup
Requires bash, curl, and jq (the scripts check and report missing ones).
The scripts live in scripts/ inside this skill directory (the folder containing this SKILL.md). Resolve all paths relative to that skill directory, not the current working directory. Run them with bash so no executable bit is required:
bash <skill-dir>/scripts/search.sh ...
If the skill directory is unknown, locate it (covers the usual harness skill roots):
find ~/.claude ~/.agents ~/.pi/agent ~/.cursor ~/.codex -path "*inciteful/scripts/search.sh" 2>/dev/null | head -1
Workflow
search.sh→ find the seed paper(s) and grab their OpenAlex IDs (or pass DOIs anywhere IDs are accepted).discover.sh→ analyze the citation network around the seeds.- Iterate: add promising papers from the results as extra seeds (comma-separated) to sharpen the network.
connect.shwhen bridging two fields,export.shto collect references.
Search papers by keyword
bash <skill-dir>/scripts/search.sh "query" [limit]
Fast title search returning id, DOI, authors, journal, and citation count.
Analyze the network around seed papers
bash <skill-dir>/scripts/discover.sh <id1[,id2,...]> [analysis] [limit] [min_year] [max_year] ["keywords"]
analysis:similar(default),important,reviews,recent,authors,upcoming,institutions,journals- Seeds accept OpenAlex IDs, bare DOIs, or DOI URLs (resolved via OpenAlex).
- The recommendation from Inciteful: add at least 3–5 seeds for a well-focused network.
- Use
keywords(e.g."heart rate variability") to filter network papers by title match when results feel off-topic.
Analysis semantics:
| Analysis | Surfaces |
|---|---|
similar |
papers citing the same literature as the seeds (biases newer) |
important |
highest PageRank in the network (biases seminal older works) |
reviews |
papers citing the most network papers (likely reviews/surveys) |
recent |
most important papers from the last 3 years |
authors / upcoming |
top authors by partial PageRank / new authors (< 10 years) |
institutions / journals |
top institutions and journals for the field |
Look up a paper
bash <skill-dir>/scripts/paper.sh <OpenAlex_ID or DOI>
Returns authors with institutions, PDF links, and samples of its references and citations.
Connect two papers (Literature Connector)
bash <skill-dir>/scripts/connect.sh <from> <to> [extend]
Finds shortest citation paths between two papers. Use extend=5 if no path is found initially (slower).
Run raw SQL on the network
bash <skill-dir>/scripts/query.sh <ids> "SELECT ..." # or a .sql file
Tables: papers(paper_id, doi, title, authors, journal, published_year, num_cited_by, num_citing, adamic_adar, cocite, page_rank, distance) and authors(paper_id, name, affiliation, partial_page_rank, ...), plus title_search('kw'). distance = 0 for seeds, 1 for their direct neighbors. Examples: papers citing 50+ network papers (reviews logic), most-cited papers excluding the seeds, author overlap between fields.
Export citations
bash <skill-dir>/scripts/export.sh <id1[,id2,...]> [bib|ris]
Tips
- Queries build a live citation graph server-side; allow 5–60 s for
discover.sh/query.sh. - For a literature review:
search.sh→discover.sh similarandimportant→ add 3–5 of the best papers as seeds →reviews+recent→export.sh ... bib. - No abstracts from Inciteful; feed IDs into the
paper-searchskill for abstracts and full metadata.