What the code does
Research-kit is the published part of the workflow behind my papers. Each tool is a single file, utilizing stdlib Python for simplicity and ease of use.
One tool compares quotes in my notes to the actual PDF text, page by page, and fails if a mismatch is found. Another tool checks for cited sentences without corresponding notes, while a third tool identifies duplicate BibTeX keys and broken DOIs. Readers can install the full Claude Code plugin or copy individual tools to enhance their research workflow. These tools check mechanics, but reading and argumentation remain with the researcher, and this is version 0.1.0, an early release.
The tools
| Stage | Tool | Checks or does |
|---|---|---|
| search | lit-search |
Queries OpenAlex and Semantic Scholar, dedupes, flags retractions, returns BibTeX |
| convert | pdf2md |
PDF to markdown with <!-- page N --> markers |
| notes | verify-quotes |
Every quote in a note matches the source page verbatim |
| review | check-claims |
Every cited sentence has a supporting note |
| review | check-bib |
Duplicate keys, missing fields, bad DOIs, cited-but-missing keys |
| build | build-paper |
Pandoc build that fails on unresolved citations or missing figures |
| analysis | new-pipeline |
Scaffolds a rerunnable Python or R pipeline |