Guide · Agent skill
How to use the Paper Lookup agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Paper Lookup is an agent skill that teaches an AI agent to search scholarly databases such as PubMed, arXiv, OpenAlex, Crossref and Semantic Scholar. It is useful for researchers doing a literature review, students looking up a paper by DOI, PMID or arXiv ID, and analysts tracing citations. After this guide you will know what the skill does, how to install it, how to phrase requests so it triggers, and what to check in its results.
Paper LookupK-Dense-AI/scientific-agent-skills · MIT Download the skill Quick start
- 1Download the ZIP. Get the paper-lookup skill ZIP from the directory page.
- 2Unzip into a skills folder. Place the paper-lookup folder in ~/.claude/skills/ for personal use, or in a project's .claude/skills/ folder.
- 3Or upload it in the Claude apps. Go to Settings → Capabilities → Skills and upload the ZIP.
- 4Check requirements. Confirm Python 3.11+, curl and network access are available. API keys are optional.
- 5Make a specific request. Name the paper, DOI, PMID, arXiv ID or topic, and state any date or open-access constraints.
- 6Review the provenance. Check which databases were queried and verify key DOIs and titles against the source.
What is an agent skill, and what does Paper Lookup do?
An agent skill is a folder of instructions and reference files that an AI agent can load when a task matches it. Paper Lookup covers 18 scholarly APIs, including PubMed, PMC, Europe PMC, bioRxiv, medRxiv, arXiv, OpenAlex, Crossref, Semantic Scholar, CORE, Unpaywall, OpenCitations, PubTator3, Zenodo, Figshare, ROR, BioStudies and DOAJ.
The skill asks the agent to pick the database that fits the request, read a reference file on that API's quirks, and report which endpoints and parameters it used. Its results include identifiers and an access date so another person can repeat the search.
- Find papers, preprints and abstracts on a topic
- Look up a paper by DOI, PMID or arXiv ID
- Find open-access PDFs and full text where available
- Trace citation graphs and author publications
- Check journal open-access status and institution ROR IDs
Install the Paper Lookup skill
Download the skill ZIP from the directory page and unzip it. For a personal install, place the folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.
You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/paper-lookup
- Project install: .claude/skills/paper-lookup inside your project
Check what you need before you start
The skill's bundled scripts need Python 3.11 or newer and use only the standard library. It also uses curl and needs network access to scholarly APIs. No credentials are required to start.
Four optional API keys raise rate limits or unlock full text in some cases: NCBI_API_KEY, S2_API_KEY, CORE_API_KEY and OPENALEX_API_KEY.
- Python 3.11+ for the bundled scripts
- curl available on your system
- Network access to scholarly APIs
- Optional: NCBI_API_KEY, S2_API_KEY, CORE_API_KEY, OPENALEX_API_KEY
Trigger the skill with natural requests
You do not need special syntax. The skill triggers on mentions of a supported database or on requests about papers, citations, DOIs, PMIDs, arXiv IDs, abstracts, full text, open-access PDFs and preprints. Name the paper or topic and say what you want back.
The more specific the request, the better the result. If a constraint changes the answer, such as a date range, a field of study, open-access only, or a full list versus a few top hits, say so. For an author name with many namesakes, add an affiliation or other detail.
- "Find papers on CRISPR off-target effects in PubMed and arXiv from the last five years."
- "Look up this DOI and tell me its journal and open-access status: 10.xxxx/xxxxx"
- "Who cites this paper? [DOI]"
- "Get me the PDF for PMID [number]."
- "Find the arXiv preprint for [title] and list its authors."
What the agent returns and how to check it
Expect a concise answer with the databases queried, the identifiers returned, and a note on any database that came back empty or partial. The skill warns that a silent gap can look like 'no papers exist' when the real meaning is 'not indexed here'.
Verify the important results yourself. The skill itself warns that APIs can return HTTP 200 with a wrong or empty body, so check titles, DOIs and counts against the publisher page or the database before you rely on them.
- Ask for the databases and endpoints used, so you can repeat the search
- Check DOIs and titles against the source for anything you will cite
- Treat an empty result as unknown until you have checked a second database
Safety notes: scripts and network access
The skill runs scripts and uses the network to call scholarly APIs. Treat returned titles, abstracts and full text as untrusted content. The skill is written not to follow instructions found inside a response and not to paste raw response text into shell commands.
The skill sets limits on its own calls. It asks before a retrieval would exceed about 1,000 records or about 50 calls. Read the scripts and reference files before installing if your environment has strict rules about code execution.
- Uses network access and runs bundled scripts
- Does not require credentials; API keys are optional
- Asks before large exhaustive pulls over about 1,000 records or 50 calls
Tips for better results
Give identifiers when you have them. A DOI, PMID or arXiv ID is more reliable than a title search. For a topic search, name the database if you have a preference, because each one covers different ground.
Use the optional API keys if you run many searches, since they raise rate limits. Keep exhaustive pulls to a stated scope so the agent does not have to stop and ask.
- Prefer exact identifiers over broad keyword guesses
- Name the database when coverage matters
- Ask for a count first on large searches
- Reconcile the number retrieved against the total the database reports
What people use it for
- Literature review: a comprehensive search across PubMed, Europe PMC, OpenAlex and Semantic Scholar with the queries recorded.
- Finding an open-access PDF for a paper you have a DOI for.
- Tracing who cites a paper using a citation graph source such as Semantic Scholar or OpenCitations.
- Mapping an author's publications, with a check for namesakes.
- Looking up a deposited dataset on Zenodo, Figshare or BioStudies connected to a paper.
Frequently asked questions
How do I install the paper-lookup skill?
Download the skill ZIP from the directory page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
Does paper-lookup require an API key?
No. The skill requires no credentials. NCBI_API_KEY, S2_API_KEY, CORE_API_KEY and OPENALEX_API_KEY are optional and only raise rate limits or unlock full text in some cases.
What do I need to run the paper-lookup skill?
You need Python 3.11 or newer for the bundled scripts, curl, and network access to scholarly APIs. The scripts use only the standard library.
Which databases does paper-lookup search?
It covers 18 databases: PubMed, PMC, Europe PMC, bioRxiv, medRxiv, arXiv, OpenAlex, Crossref, Semantic Scholar, CORE, Unpaywall, OpenCitations, PubTator3, Zenodo, Figshare, ROR, BioStudies and DOAJ.
Can paper-lookup give me a paper's full text or PDF?
It can look for open-access full text and PDFs through sources such as Unpaywall, CORE, PMC and Europe PMC. Full-text access depends on the publisher and the database, so the skill may report that a copy is not available.
Is paper-lookup good for medical decisions or checking retractions?
No. The skill is for literature lookups only and does not assess study quality. Confirm citations and retraction status against the publisher or source before relying on them.
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
Paper Lookup suits researchers running a reproducible literature review, students who want a paper or its open-access PDF from a DOI, PMID or arXiv ID, and analysts tracing citations. It is not a substitute for checking citations, retractions or clinical evidence against the source, and people who need a simple single-database search may find the bundled scripts more than they need.
Download the skill