Evaluating Paper Relevance

Screen research papers by scoring abstracts and extracting requested data.

118|12|Updated Oct 11, 2025
One-click install
npx skills add https://github.com/kthorn/research-superpower --skill evaluating-paper-relevance
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: Evaluating Paper Relevance
Source: https://github.com/kthorn/research-superpower/tree/main/skills/research/evaluating-paper-relevance
Command: npx skills add https://github.com/kthorn/research-superpower --skill evaluating-paper-relevance

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Sifting through vast search results to find papers that contain specific data or methods is a time-consuming and often imprecise task. This skill automates a two-stage screening process to efficiently identify and extract critical information, ensuring you get exactly what you need.

Core Features & Use Cases

  • Two-Stage Screening: Quickly scores abstracts for initial relevance, then performs a deep dive into full text for promising papers.
  • Integrated Access: Automatically checks ChEMBL for curated data and uses Unpaywall to find free versions of paywalled papers.
  • Structured Extraction: Extracts specific data (measurements, protocols, datasets) and organizes findings in SUMMARY.md.
  • Use Case: After a literature search, ask Claude to "evaluate these 50 papers for IC50 data." This skill will systematically screen each paper, fetch full text (even from paywalls), extract the specific data you need, and organize it in your research summary.

Quick Start

Example: Evaluate a list of papers for specific data

"Evaluate these papers for IC50 values: PMID1, PMID2, PMID3."

Frequently Asked Questions about Evaluating Paper Relevance

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I screen research papers for specific data like IC50 values or experimental protocols?

Paper screening uses a two-stage workflow: Stage 1 scores abstracts for relevance (0-10 scale), Stage 2 extracts targeted data from full texts. This skill automates both stages, fetching open-access or paywalled papers via PMC/DOI and Unpaywall, then organizes findings in a structured summary for domains like medicinal chemistry, genomics, and computational methods.

Can I automate literature review across 50+ papers without manually reading each one?

Yes. The skill systematically screens large paper sets by applying abstract scoring first, then deep-diving into full text only for promising results. It automatically retrieves papers (including paywalled ones), extracts specific measurements or datasets, and consolidates results in SUMMARY.md, eliminating manual full-text review of irrelevant papers.

What's the fastest way to extract structured data from a literature search result list?

Provide PMIDs or paper identifiers with your data request (e.g., 'evaluate these papers for IC50 values'). The skill scores abstracts first to filter candidates, then retrieves full text automatically and extracts structured data—measurements, protocols, datasets—organized by paper, reducing extraction time from hours to minutes.

How do I access paywalled research papers during literature screening?

The skill integrates Unpaywall checks and uses ChEMBL for curated data lookups. When full text isn't freely available through PMC or DOI, Unpaywall identifies legal free versions. If no free version exists, the skill documents unavailable papers so you know which require institutional access or manual retrieval.

Does this work for literature screening across different research domains?

Yes. The skill applies to literature screening across medicinal chemistry, genomics, ecology, and computational methods. Abstract scoring and data extraction logic generalizes across domains—you define the specific data or methods to search for, and the two-stage workflow adapts to your field's terminology and data types.

What happens if a paper doesn't have the data I'm looking for?

The skill scores abstracts and full texts against your criteria. Papers scoring low in Stage 1 are filtered out early. For Stage 2 deep-dives, if targeted data isn't found, the skill documents this in SUMMARY.md with reasoning—either the paper lacks the data, uses different terminology, or focuses on a different aspect, helping you identify search gaps or refine criteria.