Run systematic literature reviews from a research question
A physician-researcher built Claude Code skills that search databases, screen articles, chase citations, fetch PDFs and write the synthesis.
Done withClaude Code
- What they did
- Dalton Haslam, MD MBA, built a set of Claude Code skills backed by a Python layer that queries free literature APIs. The agent interviews you to set the PICO question, then searches PubMed, OpenAlex and Semantic Scholar. It deduplicates, screens abstracts twice, chases citations, downloads open-access PDFs, appraises quality and writes the synthesis. Six checkpoints need your approval, and you download paywalled PDFs yourself.
- How it went
- The repo includes a worked demo review on congestive heart failure care transitions, with the full output package. The source gives no timing or accuracy figures. Its disclaimer says to verify all outputs.
- Worth knowing
- A full review typically takes about 4–6 sessions. Get a free Semantic Scholar key, because the shared pool often rate-limits. The manual PDF step assumes macOS.
Try it yourself with Claude Code
I need a systematic literature review on this question: [research question]. Search [databases, e.g. PubMed, Semantic Scholar], screen results against these inclusion criteria: [criteria], follow citations from the most relevant papers, and save the PDFs you can access into a local folder. Finish by writing a synthesis document with a list of included studies, a short summary of the findings, and a log of what you excluded and why.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: github.com · Undated
Related use cases
Turn tickets into merged pull requests
Pulls tasks from GitHub Issues or Linear, then runs Claude Code or Codex in isolated Kubernetes pods to carry each one through to a PR.
Rewrite low-CTR meta titles and audit sites for SEO
An SEO agency tests Claude Code with a CMS API and Search Console to rewrite meta tags for low-CTR pages, push changes, and audit robots.txt, sitemaps and structured data.
Rebuild a math curriculum into a one-year plan with materials
Claude restructured an accelerated math curriculum into one year, writing tests, games, worksheets, a schedule, and a hosted site integrated with a teacher planner.
Run persistent financial research with an agent
LangAlpha is a Claude Code-style agent for investing research that keeps persistent workspaces and memory so analysis such as models and comps can be updated across sessions.
Custom agent
Use Cursor for academic paper writing with citations
Forum users write papers in markdown with Zotero citations and a bibtex file for Cursor to pick references from, converting to Word with pandoc.
Investigate papers overnight with a bot research team
Grok Bot roles read new papers and extract claims, rerun analyses, and scan sources, working overnight to produce briefs.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.