Skip to content
Get packs

Guide · Agent skill

How to use the Research Literature Review agent skill (research-lit)

Updated · written with AI from the project's own documentation and checked against our catalogue data

The research-lit skill is a SKILL.md-based instruction package for coding agents such as Claude Code. Once installed, it lets the agent search for academic papers on a topic, check your local PDF folders and optional reference managers, and summarize key ideas and related work. This guide covers what the skill does, how to install it, how to trigger it, which sources it uses, and where to be careful.

Research Literature Reviewwanshuiyin/Auto-claude-code-research-in-sleep · MIT

Quick start

  1. 1Download the ZIP. Download the research-lit skill ZIP from this page.
  2. 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for personal use or .claude/skills/ inside a project.
  3. 3Or upload it in the Claude apps. Go to Settings → Capabilities → Skills and upload the skill there.
  4. 4Add your papers. Place PDFs in a papers/ or literature/ folder in your project, or set a custom path under Paper Library in AGENTS.md.
  5. 5Make a request. Ask naturally, for example find papers on your topic, or run the skill with a sources directive.
  6. 6Check the results. Verify summaries and references against the original papers before using them.

What is an agent skill and what does research-lit make the agent do?

An agent skill is a folder containing a SKILL.md file that tells a compatible agent how to handle a particular kind of task. The research-lit skill is designed for searching and analyzing research papers, finding related work and summarizing key ideas.

By default the agent checks several literature sources in priority order: Zotero, Obsidian, local PDFs and web search. Local PDFs are read from the first 3 pages of up to 20 files, prioritized by filename relevance. The skill can also fetch arXiv metadata such as title, abstract and authors. Downloading arXiv PDFs is turned off by default.

  • Searches local PDF folders named papers/ or literature/ in the current project, or a custom path set in AGENTS.md
  • Checks Zotero and Obsidian through MCP servers when they are configured
  • Runs web search for arXiv, Semantic Scholar and Google Scholar results
  • Adds Semantic Scholar, DeepXiv, Exa, Gemini or OpenAlex only when you name them in the request

How do you install the research-lit skill?

Download the ZIP for the research-lit skill from this page and unzip it. You can install it for yourself across all projects or for a single project.

For a personal install, unzip the skill into ~/.claude/skills/. For a project-only install, unzip it into the .claude/skills/ folder inside your project. If you use the Claude apps instead, you can upload the skill under Settings → Capabilities → Skills.

  • Personal: ~/.claude/skills/
  • Project only: .claude/skills/ inside your project folder
  • Claude apps: Settings → Capabilities → Skills, then upload

How do you trigger the skill with natural requests?

The skill's description lists phrases that tell the agent when to use it. You can start a request with phrases such as find papers, related work, literature review, what does this paper say, or ask to understand academic papers. The agent then runs the skill with your topic as the research topic.

You can also invoke it directly with a topic and optional source directives. The source documentation shows the slash-command form, for example /research-lit "diffusion models".

  • "Find papers on contrastive learning for speech."
  • "Do a literature review on graph neural networks for drug discovery."
  • "What does this paper say about the training method?" (with a PDF in papers/)
  • "Find related work on retrieval-augmented generation, web only."
/research-lit "diffusion models" — sources: zotero, local

Which sources can you control and how?

Add a sources directive to limit where the agent searches. If you do not specify one, it uses the default set, which is every available source except Semantic Scholar, DeepXiv, Exa, Gemini and OpenAlex. Those extra sources run only when you explicitly list them.

If you ask for Zotero or Obsidian and it is not configured, the skill stops and explains how to enable it. Sources you did not request may be skipped silently.

  • sources: all (default) searches every default source
  • sources: zotero, local limits the search to Zotero and local PDFs
  • sources: web skips all local sources
  • sources: all, deepxiv adds DeepXiv to the defaults
  • arxiv download: true downloads the top relevant arXiv PDFs, up to 5 by default

What does the skill need before it runs?

The skill itself needs Claude Code or another agent that reads SKILL.md files. The optional sources need their own setup. The Zotero MCP server, Obsidian vault MCP server, Gemini MCP or CLI, and the fetcher scripts for Semantic Scholar, DeepXiv, Exa and OpenAlex are all optional. Each fetcher requires its own API key or account where applicable.

The OpenAlex fetcher needs Python with the requests package. The skill uses network access and can run scripts, so it is not a fully offline tool.

  • Claude Code or another SKILL.md-compatible agent
  • Optional: Zotero MCP server and Obsidian vault MCP server
  • Optional: Gemini MCP or CLI
  • Optional fetcher scripts for Semantic Scholar, DeepXiv, Exa and OpenAlex, each with its own API key or account where needed
  • Python with requests for the OpenAlex fetcher

What safety points should you check before using it?

The skill runs scripts and uses the network. Queries and content sent to web or AI-based sources such as Exa and Gemini may leave your machine. Do not send confidential or unpublished documents to those sources.

Summaries can be incomplete or wrong. Treat the output as a starting point and check claims and citations against the original papers before you rely on them in a bibliography or submission.

  • Do not use web or AI-based sources for confidential or unpublished material
  • Check every citation and summary against the original paper
  • Keep arXiv downloads off unless you want files saved to your library

What tips help you get better results?

Put the papers you care about in papers/ or literature/ so the agent can read them without a web search. Add a custom library path in AGENTS.md under a Paper Library heading if your collection lives elsewhere.

Name specific sources when you know where your references are, and name the optional academic databases only when you need them. Ask for a narrow question about one paper if you want a focused summary.

  • Use clear, specific topics
  • Name sources to keep the search focused
  • Use arxiv download: true only when you want PDFs saved

What people use it for

  • Starting a literature review by collecting related work from local PDFs and arXiv metadata
  • Getting a quick overview of what a single paper says
  • Searching a Zotero library or Obsidian research notes for papers on a topic
  • Finding published venue papers by adding Semantic Scholar to the sources

Frequently asked questions

Is the research-lit skill free?

The skill is released under the MIT licence. Some optional sources such as Semantic Scholar, DeepXiv, Exa and OpenAlex need their own API key or account, so check each provider's terms before using them.

What do I need to install research-lit?

You need Claude Code or another agent that reads SKILL.md files. Install it by unzipping it into ~/.claude/skills/ or a project's .claude/skills/ folder, or by uploading it under Settings → Capabilities → Skills in the Claude apps.

Does research-lit download papers automatically?

No. By default it only fetches arXiv metadata such as title, abstract and authors. Downloading up to 5 arXiv PDFs requires setting arxiv download to true.

Can the skill replace a citation manager?

No. The source says summaries may be incomplete or wrong, so it is not suitable for final citation-accurate bibliographies. Check every reference against the original paper.

Why is my request not searching Zotero?

Zotero needs the Zotero MCP server configured. If you explicitly request Zotero and it is not set up, the skill stops and explains how to enable it. Sources you do not request may be skipped silently.

How is this different from a general web search?

The skill checks your local PDFs and reference managers first, in priority order, and can add academic sources such as Semantic Scholar or OpenAlex when you name them. A plain web search does not read your own paper collection.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use research-lit if you are a graduate student or researcher starting a literature review, want quick overviews of papers, or keep your references in a local folder or Zotero. Pick something else if you need a final citation-accurate bibliography or must keep confidential documents off web and AI-based services.

Download the skill