research-papers

A plugin for managing annotated research paper collections. Works with Claude Code, Codex CLI, and Gemini CLI.

This repo supports two installation models:

• Claude Code marketplace repo via .claude-plugin/marketplace.json
• Script-based skill installer for Codex CLI, Claude Code, and Gemini CLI via scripts/install_skills.py

What it does

This plugin provides skills for retrieving, reading, and annotating scientific papers into a structured, cross-referenced collection. Each paper gets:

• notes.md - Implementation-focused extraction (equations, parameters, algorithms)
• description.md - Three-sentence summary
• abstract.md - Verbatim abstract plus interpretation
• citations.md - Full reference list plus key citations for follow-up
• claims.yaml / justifications.yaml / stances.yaml - Optional propstore-backed argument artifacts
• papers/index.md - Auto-generated list of paper directories in the collection

Propstore

Most of the propositional-knowledge pipeline skills (source-bootstrap, author-context, register-concepts, extract-claims, enrich-claims, extract-justifications, extract-stances, register-predicates, author-rules, author-lifting-rules, source-promote, ingest-collection, ingest-new-papers, paper-process) invoke the pks CLI from propstore. Install it once with:

uv tool install git+https://github.com/ctoth/propstore

This provides the pks executable on your PATH. Skills that do not touch the propstore source-branch layer (paper-retriever, paper-retriever-scihub, paper-retriever-institutional, paper-retriever-bookshare, paper-reader, process-new-papers, reconcile, lint-paper, tag-papers, verify-citations, research) work without it.

Skills

Retrieval and reading

Skill	Description
paper-retriever	Retrieve a scientific paper PDF from arxiv, DOI, open repositories, publisher HTML, or an enabled access backend
paper-retriever-scihub	Paywalled-access backend used by paper-retriever when open-access routes fail
paper-retriever-institutional	Paywalled-access backend for configured institutional or library-proxy access
paper-retriever-bookshare	Credentialed Bookshare backend for retrieving books as EPUBs
paper-reader	Read a paper and extract structured notes (handles small/medium/large papers)
paper-process	Full per-paper flow: retrieve, read, and run the propstore ingestion pipeline
process-new-papers	Batch paper-reader over every unprocessed PDF in papers/ root

Propstore ingestion