ai4scholar-pdf-fulltext-reading

name	ai4scholar-pdf-fulltext-reading
description	Use AI4Scholar to download open-access PDFs or read full text from Semantic Scholar, arXiv, bioRxiv, medRxiv, and DOI-based sources when access permits. Trigger for reading papers, extracting methods/results, PDF download, full-text summaries, or DOI full-text retrieval.

AI4Scholar PDF And Full-Text Reading

Use this skill when the user wants to read or summarize paper full text, download PDFs, inspect methods/results, or extract evidence from an article.

Core Rules

• Prefer legal open-access routes first.
• DOI download for paywalled papers works only when the running environment has institutional access.
• If extraction is poor, say so and do not infer missing methods or results.
• Do not summarize a full paper from title/abstract alone unless explicitly labeled as abstract-only.

Tools From The Source Guide

Tool	Use
download_semantic	Get Semantic Scholar open-access PDF link
read_semantic_paper	Download and extract Semantic Scholar full text
download_arxiv	Get arXiv PDF link
read_arxiv_paper	Download and extract arXiv full text
download_biorxiv	Get bioRxiv PDF link
read_biorxiv_paper	Download and extract bioRxiv full text
download_medrxiv	Get medRxiv PDF link
read_medrxiv_paper	Download and extract medRxiv full text
download_by_doi	Download PDF by DOI when access permits
read_by_doi	Download and extract full text by DOI when access permits

Reading Workflow

1. Verify the paper identifier.
2. Try open-access source in this order:
   • arXiv/bioRxiv/medRxiv if the identifier is a preprint ID,
   • Semantic Scholar open-access PDF,
   • DOI access when permitted.
3. Extract and segment:
   • Abstract,
   • Introduction,
   • Theory/background,
   • Method,
   • Data/sample,
   • Measures/stimuli,
   • Analysis,
   • Results,
   • Discussion/limitations.
4. For empirical papers, record exact design, sample, variables, model, and key statistics.
5. Mark any unreadable sections.

Evidence Extraction Template

Paper:
Identifier:
Access route:
Extraction quality: good / partial / poor

Research question:
Theory/mechanism:
Data/sample:
Method/design:
Variables/measures:
Key results:
Limitations:
Reusable ideas for my project:
Sections that need manual PDF reading:

Beginner Prompt

Use AI4Scholar to try to read the full text of this paper: [DOI/arXiv/PMID/title].
Extract research question, mechanism, sample, experiment/data design, variables, model, key results, and limitations.
If only the abstract is available or PDF extraction fails, say so clearly and do not guess.

Fallbacks

• Use the local pdf or markitdown skill for downloaded PDFs.
• Use publisher HTML when PDF extraction is bad.
• For scanned PDFs, tell the user OCR is required.