Skip to content
Get packs

Guide · Agent skill

How to use the Research Lookup agent skill for manuscripts

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

The research-lookup skill is a set of instructions and a bundled Python script that gives an AI agent a repeatable way to gather scholarly literature for a scientific manuscript or research brief. This guide explains what it does, how to install it from a ZIP file, how to ask for it in plain language, and what to check before you cite anything it returns. By the end you will be able to run it on a research question and review the packet it produces.

Research Lookup for ManuscriptsK-Dense-AI/scientific-agent-skills · MIT

Quick start

  1. 1Download and unzip the skill. Download the research-lookup ZIP from this page and unzip it into ~/.claude/skills/, or into the project's .claude/skills/ folder.
  2. 2Set up credentials. Provide PARALLEL_API_KEY or log in with the parallel-web-tools CLI. Add OPENROUTER_API_KEY only if you plan to use Perplexity.
  3. 3Write a context file. Create a JSON file with your research question, study type, population, exposure, comparator, outcomes, and field. Leave out anything you do not know.
  4. 4Ask for the packet explicitly. Request literature or a manuscript research packet in plain language so the agent activates the skill.
  5. 5Run the academic pipeline. Run the research_lookup.py script with --academic, --context-file, and --packet-dir, as shown in the workflow.
  6. 6Review the packet. Open packet.md and the claim-source map, then check each source's identity, support for your claim, and retraction status.

What the research-lookup skill makes the agent do

The skill activates only when you explicitly ask for literature, references, background evidence, competing findings, or a manuscript research packet. It then tells the agent to run a bundled script that searches the web through Parallel's search service and retrieves the text of promising sources.

Rather than returning a loose list of links, the skill organizes results into a research packet. The packet includes a reference ledger and a claim-to-source style evidence summary. A reference counts as verified only when its text was successfully retrieved. Retrieval does not confirm that the citation matches the paper you intend, that the source supports your claim, or that it has not been corrected or retracted; you still need to check those yourself.

  • Default academic run: Parallel Search for candidates, then Parallel Extract for source text
  • Explicit deep or exhaustive reports: Parallel Research, selected only when you ask for it
  • Optional Parallel Chat or Perplexity via OpenRouter, only when explicitly requested or enabled as a fallback

How to install the research-lookup skill

Download the ZIP of the skill from this page. You can install it in one of two ways, depending on where you use your agent.

For Claude Code, unzip the download into your personal skills folder at ~/.claude/skills/ so it is available in all projects. To limit it to one project, unzip it into that project's .claude/skills/ folder instead. In the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.

  • Personal install: ~/.claude/skills/
  • Single-project install: the project's .claude/skills/ folder
  • Claude apps: Settings → Capabilities → Skills, then upload the ZIP

How to trigger the skill with natural requests

Ask for the work explicitly. The skill is designed not to run on casual factual questions, so phrases that name literature, references, or a manuscript packet are the most reliable. Give the agent your research question and any study details you have, and do not make up missing details.

The academic run in the skill's workflow looks like this. Replace the placeholder query with your own research question and point --context-file at a JSON file describing your study. Note that this run uses a paid search service, so review its scope first.

  • Example request: Compile a research packet on how intervention X affects outcome Y, with recent primary studies and any null or contradictory findings.
  • Example request: Find background references and competing findings for the introduction of my manuscript on this topic.
  • Example request: Build a manuscript research packet with DOIs and PMIDs I can review, using only the study details in my context file.
python skills/research-lookup/scripts/research_lookup.py "Evidence relevant to the manuscript's research question" --academic --target-references 60 --context-file manuscript-context.json --packet-dir sources/manuscript-research --json

What you need before running it

The skill requires Python 3.10 or newer and network access. Search and extraction need the parallel-web-tools command line tool, which can authenticate with a PARALLEL_API_KEY environment variable or with a stored CLI login. Never put API keys in command arguments.

Two other keys are optional. Explicit Parallel Chat needs the requests library and PARALLEL_API_KEY. Optional Perplexity use through OpenRouter needs the requests library and OPENROUTER_API_KEY. You only need these if you select those backends.

  • Python 3.10+
  • Network access
  • parallel-web-tools CLI or PARALLEL_API_KEY
  • requests and PARALLEL_API_KEY for explicit Chat
  • OPENROUTER_API_KEY for explicit Perplexity

Safety notes: scripts and network use

This skill runs scripts and uses the network. Query text and context you provide are sent to Parallel, and sent to OpenRouter only when Perplexity is selected or enabled. Do not use it for private or unpublished data, because that material would leave your machine.

The skill also treats web content returned by searches as untrusted data, not as instructions. It is not designed to guarantee completeness, so it is not a replacement for a systematic review. For PRISMA-style work, the skill points you to a different skill, literature-review.

  • Do not include unpublished study data or confidential material in queries or context files
  • Review the scope of a paid run before starting it
  • Use literature-review for systematic reviews and protocols

Tips for reviewing the research packet

The output directory contains several files. Start with packet.md for a readable overview, then open references.bib and the claim-source-map to check each claim against its excerpt. The coverage and search-ledger files show how many references were targeted, which filters and searches ran, and any errors or unresolved URLs.

Before citing anything, confirm the bibliographic identity of each paper, read the excerpt against the claim you want to support, and check correction and retraction status at the publisher or an authoritative index.

  • Read packet.md first, then the claim-source map
  • Check coverage.json for target shortfall and errors
  • Verify every DOI or PMID and retraction status yourself
  • Treat search-only records as unverified

What people use it for

  • Drafting the introduction of a manuscript by finding recent primary studies and reviews on a topic
  • Collecting methods precedent for protocols, measures, or analyses before writing the methods rationale
  • Surfacing contradictory or null findings to inform the discussion section of a paper

Frequently asked questions

How do I install the research-lookup skill?

Download the ZIP from this page and unzip it into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ folder for one project. In the Claude apps, you can instead upload the ZIP under Settings, then Capabilities, then Skills.

What do I need to run research-lookup?

You need Python 3.10 or newer, network access, and either the parallel-web-tools CLI or a PARALLEL_API_KEY. Explicit Parallel Chat additionally needs the requests library, and optional Perplexity use needs OPENROUTER_API_KEY.

Is the research-lookup skill free?

The skill is released under the MIT licence. The search services it calls are external, and the skill's own workflow notes that its academic run uses a paid service, so check the cost of your Parallel account before running it.

Does research-lookup guarantee that citations are real and correct?

No. A reference is counted as verified only when its text was retrieved, which does not confirm bibliographic identity, claim support, peer review, or retraction status. You need to check each source yourself.

Can I use research-lookup for a systematic review?

No, the skill says it cannot guarantee completeness. For PRISMA-style work with protocols, database-specific searches, and screening, the skill points you to the literature-review skill instead.

Is it safe to use with unpublished data?

Avoid it for private or unpublished material. Query text and context are sent to Parallel, and to OpenRouter if Perplexity is used.

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-lookup if you need a starting set of recent papers, reviews, and competing findings for a manuscript introduction or discussion, and you are prepared to verify every source yourself. Pick a different tool if you need a systematic review with protocols and screening, if you expect citations to be verified automatically, or if your material is confidential.

Download the skill

Find exactly what
you're looking for