Which online databases does it query?

It primarily utilizes the arXiv API for retrieval and Semantic Scholar (via the r.jina.ai proxy) for citation snowballing and metadata backfilling.

What is the primary goal of the Literature Engineer skill?

Its goal is to build a large, verifiable candidate pool of at least 200 research papers with stable identifiers (DOIs/arXiv IDs) to serve as the evidence foundation for research projects.

Can I use this skill without an active internet connection?

Yes, it supports offline-first workflows by merging and normalizing existing local exports in .bib, .csv, .json, or .jsonl formats.

What formats are the results delivered in?

The skill generates a structured papers_raw.jsonl for programmatic use, a CSV for human review, and a Markdown retrieval report summarizing the data collection stats.

How does it prevent AI hallucinations?

The skill follows an 'evidence-first' methodology. If it cannot find or verify a paper's provenance or ID, it is programmed to block the process and request more data rather than fabricating titles.

Literature Engineer

Name: Literature Engineer
Author: WILLOSCAR

byWILLOSCAR

•

数据科学与机器学习

Automates the gathering, normalization, and verification of large-scale research paper datasets for evidence-based surveys.

The Literature Engineer skill is a robust research utility designed to build high-volume, verifiable candidate pools of academic papers. By employing a multi-route approach—integrating offline imports, online API retrieval from sources like arXiv, and citation snowballing—it ensures researchers have a foundation of at least 200 papers before beginning a survey. It prioritizes data integrity by enforcing stable identifiers (DOIs/arXiv IDs) and detailed provenance tracking, effectively preventing AI hallucinations by requiring verifiable evidence before any drafting or synthesis occurs.

主要功能

01Evidence-first guardrails that block fabrication of non-existent papers

0289 GitHub stars

03Support for large-scale candidate pools (target 200+) with structured JSONL output

04Multi-route expansion via offline files, online APIs, and citation snowballing

05Automated retrieval reports detailing coverage, provenance, and metadata health

06Metadata normalization and deduplication using stable IDs (DOI, arXiv ID)

使用场景

01Automated discovery of seminal works via forward and backward citation snowballing

02Building comprehensive bibliographies for systematic literature reviews or whitepapers

03Merging and deduplicating messy paper exports from multiple academic databases

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add willoscar/research-units-pipeline-skills literature-engineer

For use in Claude.ai and ChatGPT

Download Skill