luwill/research-skills

What it solves

This project provides a suite of specialized "skills" for Claude Code to automate and enhance academic research workflows. It addresses common pain points in scholarly work, such as the tedious process of exhaustive literature searching, the risk of AI hallucinations in citations, the difficulty of creating high-fidelity academic presentations, and the struggle of drafting structured research proposals and medical imaging reviews.

How it works

Each skill is a modular extension that can be installed into the Claude Code skills directory. They utilize a combination of LLM orchestration, specialized Python/TypeScript scripts, and integrations with academic data sources (like arXiv, PubMed, and Zotero) to execute complex research tasks:

• Lit Search: Uses a deterministic pipeline with multi-source pagination and citation closure to ensure high recall and provide a measurable coverage report.
• Scholar Slides: Focuses on source fidelity, using KaTeX for equations and native OOXML for tables to create editable PPTX or PDF decks.
• Paper Slide Deck: Uses the Gemini API to generate stylized, AI-generated images of slides for visual sharing.
• Research Proposal: Employs a 5-phase "write-with-verify" workflow to draft proposals with verified citations.
• Medical Imaging Review: Implements an 8-phase workflow specifically for medical imaging AI, incorporating reporting standards like PRISMA and an executable audit script to ensure citation integrity.

Who it’s for

Academic researchers, PhD students, and scholars (particularly those in STEM and medical imaging AI) who use Claude Code and want to automate their literature reviews, proposal writing, and presentation creation while maintaining strict academic integrity.

Highlights

• Citation Integrity: Multiple skills include hard gates and verification protocols to prevent AI-fabricated references.
• Measurable Recall: The literature search tool provides a coverage report and saturation curves rather than just a list of relevant papers.
• Fidelity-First Presentations: Scholar Slides ensures equations and data remain as vector text/tables rather than rasterized images.
• Structured Workflows: Provides multi-phase pipelines (e.g., 8-phase for medical reviews) to move from initial capture to final submission prep.
• Broad Academic Support: Includes bilingual support (English/Chinese) and integration with Zotero, PubMed, and arXiv.