Guide · Agent skill
How to use the nature-academic-search agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
nature-academic-search is an agent skill for academic literature work. It gives an AI agent a structured playbook for searching sources such as CrossRef, PubMed, arXiv, Scopus and ScienceDirect, checking citations, building PubMed MeSH strategies and running citation-impact audits. This guide explains what an agent skill is, how to install this one, how to trigger it with plain requests, and what it needs before it works.
Academic Search RouterYuan1z0825/nature-skills · Apache-2.0 Download the skill Quick start
- 1Download the ZIP. Download the skill ZIP from the skill page on this site.
- 2Place or upload the skill. Unzip into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
- 3Prepare the environment. If you need the MCP server, install the Python packages listed in mcp-server/requirements.txt and have your PubMed email ready for the installer.
- 4Add API keys only if needed. Configure Scopus and ScienceDirect keys through pybliometrics only if you need those sources.
- 5Ask in plain language. Describe your topic, time range, DOI or document and the output you want.
- 6Review the output. Check the deduplicated results and the list of unverifiable entries against the original sources.
What an agent skill is and what this one does
An agent skill is a folder of instructions that an AI agent loads when a task matches it. This skill's entry file, SKILL.md, acts as a router: it tells the agent which guidance to read for the task at hand, so it does not load everything at once.
Once installed, the agent can search literature across multiple sources, check whether citations in a document are real and correctly formatted, build MeSH-assisted PubMed search strategies, convert reference files (.nbib, .ris, .bib), and run strict external-citation audits. Its guidance covers workflows for multi-source search, citation verification, MeSH strategy, citation file management, reference management and strict other-citation impact audits.
- Multi-source literature search
- Citation verification from a document
- PubMed MeSH and search strategy building
- Conversion and management of .nbib, .ris and .bib files
- Strict external-citation audits and citer profiles
Install the skill from the ZIP or the Claude apps
Option one for Claude Code: download the ZIP from the skill page, unzip it, and place the skill folder in ~/.claude/skills/ for your user account. To limit it to one project, place it in that project's .claude/skills/ folder instead.
Option two for the Claude apps: download the ZIP and upload it in the Claude apps under Settings → Capabilities → Skills.
- Personal use across projects: ~/.claude/skills/
- Single project only: the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
~/.claude/skills/Check the requirements before you rely on it
The skill's core search and citation guidance runs as instructions, but the package also includes scripts and an optional MCP server that supplies live database tools. The MCP server needs Python with the packages listed in mcp-server/requirements.txt, and Claude Code or a compatible agent to run it.
The server requires a PubMed email address, which NCBI asks for. You pass it to the installer. An NCBI API key is optional and raises rate limits. Scopus and ScienceDirect require API keys, configured through the pybliometrics configuration file, which by default is ~/.config/pybliometrics.cfg. Without those keys, the Scopus and ScienceDirect tools will not work, though CrossRef, PubMed and arXiv remain available.
- Python with packages from mcp-server/requirements.txt
- Claude Code or a compatible agent for the MCP server
- PubMed email address (passed to the installer)
- Optional NCBI API key for higher rate limits
- Scopus and ScienceDirect API keys, if you need those sources
PUBMED_EMAILTrigger the skill with natural requests
You do not need special syntax. Describe the literature task in plain language, and the agent will match it to the skill. The skill's description lists its trigger terms, including Chinese phrases such as 文献检索 (literature search), 引文核对 (citation checking), 参考文献管理 (reference management) and 严格他引 (strict other-citation audit).
Give the agent what it needs to work: the topic, keywords, time range, journal scope or the target DOI. For an audit, also state how you want to exclude self-citations, co-author or team citations, and collaboration-network citations. Results are more useful when you name the output you want, such as a deduplicated table, a citation style or an export file.
- "Find 20 recent Nature, Science or Cell papers on this topic."
- "Convert these DOIs to Nature-style citations and export RIS."
- "Add MeSH terms and synonyms to my PubMed search."
- "Check the strict external citations for this paper and list influential citers."
Use it for citation checks and file conversion
To check a reference list, give the agent the references from your document. The skill is designed to verify whether citations are real and correctly formatted, and to report entries it cannot verify. It can also convert complete reference records between .nbib, .ris and .bib formats. For a conversion that uses records you already supplied, the skill does not require source lookup or network checks.
Expect a result that lists the deduplicated records with title, authors, year, journal, DOI and source, along with any entries that could not be verified. Review those entries yourself before you submit anything.
- Verify references extracted from a document
- Convert .nbib, .ris or .bib files with complete records
- Look up a paper by DOI, PMID or arXiv ID
- Format citations in styles such as APA, Nature, IEEE and Vancouver
Review results and understand the limits
Database access can be rate-limited or incomplete, so search results can miss papers or contain errors. Treat the output as a starting point and check important entries against the original source. If a tool fails, the skill is designed to report the failure and continue with the other tools, and to broaden search terms when there are no results.
The skill does not translate papers or draft manuscript prose. Use a different tool for those tasks.
- Check results against the original database records
- Do not rely on unverified entries without review
- Use a different tool for translation or manuscript writing
Safety: scripts and network access
This skill runs scripts and uses the network. Its scripts include a format converter, a preflight check and a no-MCP fallback discovery search. It queries external scholarly databases during search and verification.
Read the included files before installing if you need to confirm what runs on your machine. Be deliberate about which API keys you configure, since they give the skill access to your institutional or account-level services.
- Runs scripts on your machine
- Sends queries over the network to scholarly databases
- Only configure the API keys you intend to use
What people use it for
- Running a literature review across CrossRef, PubMed and arXiv with a deduplicated table
- Checking that references in a student paper are real and correctly formatted
- Building a PubMed MeSH search strategy with synonyms for an information-specialist brief
- Converting a .nbib export into .ris or .bib for a reference manager
- Auditing strict external citations of a target paper and summarising how selected citers cited it
Frequently asked questions
How do I install the nature-academic-search skill?
Download the ZIP from the skill page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one project. In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.
Is nature-academic-search free to use?
The skill is released under the Apache-2.0 licence. Some sources require API keys, such as Scopus and ScienceDirect, and those keys come from the providers, not from this skill.
What do I need to run the MCP server?
You need Python with the packages in mcp-server/requirements.txt, Claude Code or a compatible agent, and a PubMed email address that you pass to the installer. An NCBI API key is optional.
Does it work without Scopus or ScienceDirect access?
Yes for CrossRef, PubMed and arXiv. The Scopus and ScienceDirect tools need API keys configured through pybliometrics, so without them those sources will not return results.
Can it check whether my citations are real?
That is one of its purposes. It can verify citations extracted from a document and report entries it cannot verify, but you should still review the flagged and unverified items yourself.
How is it different from a general AI chat or a reference manager?
It gives the agent a structured playbook for searching multiple scholarly databases, building MeSH strategies and converting reference files. It does not translate papers or write manuscript prose, and a reference manager is still the place to store your library.
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
This skill suits researchers running multi-database literature reviews, students checking references, and librarians building PubMed search strategies, as long as they review results themselves. Skip it if you need translation or manuscript drafting, or if you need Scopus or ScienceDirect and cannot get the required API keys.
Download the skill