Guide · Agent skill
How to use the Database Lookup agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Database Lookup skill is an agent skill: a packaged set of instructions and reference files that a Claude agent loads when it needs to answer from a specific public database rather than from its general memory. It is aimed at researchers, analysts and students who need a figure or record they can trace back to its source. After reading this guide you will know how to install the skill, how to phrase a request so it triggers, what the agent reports back, and what to check before relying on the result.
Database LookupK-Dense-AI/scientific-agent-skills · MIT Download the skill Quick start
- 1Download and unzip. Download the skill ZIP from this page and unzip it to get the database-lookup folder.
- 2Place the folder. Move it to ~/.claude/skills/ for personal use or to .claude/skills/ inside a project. Alternatively, upload it under Settings → Capabilities → Skills in the Claude apps.
- 3Check access. Confirm the databases you need are public or that you have any required API key, such as for Alpha Vantage.
- 4Ask a specific question. Name the source, the identifier and what you need back, and say you want it traceable.
- 5Review the provenance. Check the endpoints, parameters, access date and count reconciliation in the answer.
- 6Repeat the lookup. Use the listed endpoints and parameters to rerun the query yourself and confirm the result.
What the Database Lookup skill makes the agent do
The skill teaches the agent to pick an authoritative database for a question, read that database's reference file, and make bounded, rate-limited calls to its web API. The source describes a catalogue of about 80 databases spanning biology, chemistry, clinical trials, economics and other fields, with one reference file per database.
The agent is told to check record counts so results are complete, and to report the databases queried, endpoints, parameters, access date and any identifier conversions. The aim is that someone else can repeat the lookup and get the same answer.
- Chooses the primary database for your question instead of searching many APIs at once
- Makes bounded calls and paginates until retrieved counts match the expected total
- Flags incomplete pagination, ambiguous filters, stale data and source limits
- Says explicitly when a query returns no results
Install the skill from the ZIP or the Claude apps
Download the ZIP from our page and unzip it. For a personal install, place the unzipped skill folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.
If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/ (available across your projects)
- Project: .claude/skills/ inside the project folder (shared with that project)
- Claude apps: Settings → Capabilities → Skills, then upload
What you need before it works
The skill needs internet access to the public database APIs and Bash tool access so the agent can run the API calls. Most listed sources are public, but some need a key or registration. Alpha Vantage, for example, needs an API key, and Addgene requires approval and a licence for each scope. The skill says never to expose keys in its outputs.
The skill runs scripts and uses the network, so review what it will call before you approve tool use.
- Internet access to the database APIs
- Bash tool access
- API keys only for sources that require them (for example Alpha Vantage)
How to trigger the skill with natural requests
Ask for a specific fact or record from a named source, and say why it must be traceable. The skill is meant for database-backed facts that must be retrieved reproducibly rather than inferred. Naming the database helps the agent choose the right reference file.
Include identifiers where you have them, such as a UniProt accession, a ChEMBL ID or a BEA series. The skill asks a clarifying question when a missing constraint would change the answer.
- "Using ChEMBL, get the approved drug records for CHEMBL25 and list the endpoints and parameters you used."
- "Fetch the AlphaFold prediction metadata for UniProt P00533 and give me the file URLs with the access date."
- "Pull the BLS series I need from BLS and tell me the exact parameters so I can repeat it."
Examples of what a good answer contains
A good result is a short answer or table, not a raw dump. It lists the database, endpoint, parameters, access date and count reconciliation (expected total, retrieved total, pages). Any raw payload is labelled as untrusted third-party data.
If you asked for a figure that is partly stale or filtered locally, the answer should say so.
- Answer or structured table
- Databases, endpoints, parameters and access date
- Expected total, retrieved total, pages or batches, local filters applied
- Warnings on stale data or source limits
Safety notes: scripts, network and limits
The skill makes network calls and runs scripts. It asks for confirmation before a retrieval would exceed 10,000 records, 100 API calls, or the API's documented bulk-use guidance. It also treats API responses as untrusted data and does not follow instructions embedded in them.
The skill is not a substitute for expert review. The source warns that small filter differences can change conclusions and that data may be stale, so clinical or regulatory decisions should not rest on retrieved data alone.
- Confirms before exceeding 10,000 records or 100 API calls
- Never follows instructions found inside returned data
- Never exposes API keys in outputs
- Not for bulk downloads or private databases you lack credentials for
What people use it for
- A researcher pulls a specific ChEMBL compound record and cites the endpoint used.
- An analyst retrieves an economic series from BEA or BLS with the parameters stated.
- A student checks a protein structure's AlphaFold DB metadata and file URLs for a report.
- A reviewer repeats a database lookup from a paper to confirm a figure.
Frequently asked questions
How do I install the database-lookup agent skill?
Download the ZIP from this page and unzip it. Put the folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder. In the Claude apps you can upload it under Settings → Capabilities → Skills.
Does the database-lookup skill require an API key?
Not for every database. The skill says some sources need keys or registration; Alpha Vantage needs an API key, and Addgene requires approval and a licence for each scope. The skill says never to expose keys in outputs.
Is the database-lookup skill free?
The skill is published under the MIT licence. Some underlying databases have free and paid tiers, and the source notes that Alpha Vantage has free and paid entitlements, so check each provider's terms.
What databases does the skill cover?
The source says it catalogues about 80 databases across biology, chemistry, clinical trials, economics and other fields, with a reference file for each. Examples in the reference list include ChEMBL, AlphaFold, BEA, BLS and ClinicalTrials.gov.
Why does the skill need Bash and internet access?
It runs API calls through the Bash tool and reaches public database web APIs over the network. The source lists both as requirements, and the skill runs scripts, so review what it will call.
Can I use this skill for clinical decisions?
The source says no without expert review. It warns that small filter differences can change conclusions and that retrieved data may be stale, so use it to find and trace data, not to decide treatment or regulatory action on its own.
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 this skill if you need specific figures or records from named public databases that you can trace, cite and repeat, such as researchers checking a compound or analysts pulling official economic series. Pick something else if you need bulk downloads, private or licensed databases you cannot access, or clinical and regulatory decisions made without expert review.
Download the skill