Academic Research Skill

Literature search, citation verification and structured notes for agents that write with references.

Try in Runcell
Made by
Runcell team
Category
Research
Format
SKILL.md
Works with
Claude Code, Codex, Cursor, Runcell (Jupyter)

Showcase

Illustrative output. The real run happens in your notebook.

references.bib
  1. [1]Attention is all you needarXiv:1706.03762 ✓
  2. [2]Deep residual learning for image recognitiondoi:10.1109/CVPR.2016.90 ✓
  3. [3]Language models are few-shot learnersarXiv:2005.14165 ✓
Every reference resolves to a DOI or arXiv ID before it is used.
Citation check
3 references verified via Crossref / arXiv
1 preprint flagged (no journal version found)
0 unresolved
BibTeX written: references.bib
The citation check runs before drafting starts.

Description

Academic work has rules that general assistants break: every claim needs a source, sources must exist, and quotes must be exact. This skill encodes those rules so an agent searches the literature, keeps a reference list with identifiers, and never cites a paper it has not seen.

It covers the loop from question to draft: find papers, extract methods and results into structured notes, compare them, and write sections with citations that resolve to a DOI or arXiv ID. In Jupyter the extracted numbers live in cells, so a comparison table is data, not prose.

  • Searches arXiv, PubMed, Semantic Scholar and Crossref through available tools
  • Verifies every citation against a DOI, arXiv ID or PubMed ID before using it
  • Extracts methods, datasets, metrics and results into a structured note per paper
  • Builds comparison tables across papers
  • Drafts related-work and background sections with in-text citations and a BibTeX file
  • Flags retracted, preprint-only or unverifiable sources
Editor's Recommendation

Recommended by

Runcell team

Why we built this skill

We built this after one too many invented citations. The rule inside is blunt: no identifier, no citation. It makes the agent slower and much more useful. The structured notes per paper turned out to be the surprise benefit, because a comparison table falls out of them for free.

What you get

  • A verified reference list with identifiers and BibTeX
  • Per-paper structured notes
  • A comparison table of methods and results
  • Draft sections with citations that resolve

Use it when

  • Related-work sections and literature reviews
  • Checking that a manuscript's references are real and correctly attributed
  • Building a reading map for a new topic
  • Graduate students and researchers who write in LaTeX or Markdown

FAQ

What are academic research skills for Claude Code?

They are skill files that give Claude Code a procedure for scholarly work: how to search databases, how to verify citations, how to take structured notes, and how to draft with references. The key difference from normal prompting is citation verification. The agent must resolve a paper to an identifier before citing it.

Does it work with Codex and Cursor?

Yes. The skill uses the Agent Skills format, which Claude Code, Codex, Cursor and Runcell all read. Database access depends on which fetch or search tools your agent has.

Can it prevent hallucinated citations?

It reduces them sharply. The rule is simple: no identifier, no citation. Every reference must resolve through Crossref, arXiv or PubMed, and the agent records the lookup. It cannot verify content behind paywalls it cannot read, and it says so.

How does it fit with the deep research skill?

Deep research is for open questions across the web. Academic research is for the scholarly literature with formal citation rules. Many projects use both: deep research to map a topic, academic research to build the reference list.

Related Skills

View all
Research

Deep Research Skill

Turn a question into a multi-step research run with sources, cross-checks and a cited report.

Runcell teamFree
Publication-style dose-response figureVisualization

DataViz Skill

Chart choice, consistent styling and journal-ready export for matplotlib, seaborn and plotly.

Runcell teamFree
Statistics

Statistical Test Selection Skill

From question and data to the right test, with assumptions checked and results reported correctly.

Runcell teamFree

Information

Format
SKILL.md (Agent Skills)
Works with
Claude Code, Codex, Cursor, Runcell (Jupyter)
Runs in
A live Jupyter kernel in Runcell, or your terminal agent
Maintained by
Runcell team
Last updated
Sep 20, 2026
Price
Free

Run this skill in your notebook

Runcell is a Jupyter-native AI agent that executes every step in a live kernel. Results are cells you can inspect and rerun.

Try in Runcell