Generate personalized literature reviews from a one-line topic
A researcher built a literature-review Skill for an agent framework that outputs Markdown, Word and PDF reviews with GB/T 7714 references carrying DOIs or URLs, used by their team.
Done with a custom-built or unnamed agent
- What they did
- The skill goes into an agent framework and takes a one-line topic. It first sets boundaries: four layers of keywords, a time-sensitivity level, and routing to arXiv or PubMed by field. Then it runs ten scripts against OpenAlex, Semantic Scholar, CrossRef, Unpaywall, arXiv and PubMed in five search rounds. Titles and DOIs are used to remove duplicates, and metadata comes from the APIs, not from memory.
- How it went
- The author has used and refined it with their team for over six months, and the students like it. Each reference carries a DOI or URL. The full post is paywalled, so later steps aren't visible.
- Worth knowing
- The online Feishu version works worse than the local one, though it is easier to share. The skill file itself is at the end of the paywalled post.
Try it yourself with Claude Code
Build me a reusable skill that takes a one-line topic like [example topic] and writes a literature review. It should find real papers, write the review in [language], and cite them in [citation style, e.g. APA or GB/T 7714] with a DOI or URL for every reference. Export it as Markdown, Word and PDF, and drop any reference you can't verify. Test it on [topic] and stop when all three files are saved.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: wangshuyi.substack.com · Oct 26, 2026 · Submitted by from search gap: 学术
Related use cases
Find connections across 100 non-fiction books
Claude Code browses a personal library with CLI tools and surfaces trails of related excerpts across books.
Summarize 20 articles into a cited document
An agent extracted key insights from 20 articles, found common themes and produced a cited summary; the write-up estimates 18 minutes versus hours by hand.
Let Cursor maintain an Obsidian notes vault
Cursor reviews vault structure, suggests connections, answers questions about notes and writes a daily done note documenting other agent sessions.
Research on X, summarize YouTube transcripts, build decks
@Tape_Vector uses Comet for X research, summarizing YouTube video transcripts and building presentations.
Run research workflows with specialized bot agents
A setup of 10 specialized Grok Bot agents includes a Research agent that finds sources and data and a Deep Research agent that compares sources into insights.
Orchestrate academic research and writing workflows
Researchers use Claude Code as an orchestrator for research workflows and as their main tool for writing papers and theses in LaTeX or Markdown.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.