gs-fulltext

name	gs-fulltext
description	Get full-text access links for a Google Scholar paper - PDF, DOI, Sci-Hub, and publisher links. Use when user wants to read or download a paper's full text.
argument-hint	[data-cid or result number from previous search]
version	0.1.0

Google Scholar Full Text Access

Resolve and present all full-text access options for a paper found in Google Scholar search results.

Arguments

$ARGUMENTS can be:

• A data-cid from a previous search result
• A result number (e.g., 3) referring to a previous search result

Prerequisites

This skill works best after a previous gs-search or gs-advanced-search call, which already extracts fullTextUrl and href for each result. If those are available, use them directly.

Steps

Step 1: Collect links from search results

From the previous search result, extract these fields for the target paper:

• fullTextUrl: direct PDF/HTML link (from .gs_ggs a)
• href: link to the paper's publisher page (from .gs_rt a)
• dataCid: cluster ID

If no previous search context is available, search for the paper first using gs-search.

Step 2: Resolve additional links (evaluate_script)

Navigate to the Google Scholar results page (if not already there), then extract full-text links for the specific paper:

async () => {
  const cid = "DATA_CID_HERE";

  // Find the result item by data-cid
  const item = document.querySelector(`.gs_r.gs_or.gs_scl[data-cid="${cid}"]`);
  if (!item) return { error: 'not_found', message: 'Paper not found on current page. Try searching again.' };

  const titleEl = item.querySelector('.gs_rt a');
  const title = titleEl?.textContent?.trim() || item.querySelector('.gs_rt')?.textContent?.trim() || '';
  const paperUrl = titleEl?.href || '';

  // Full-text PDF/HTML link (shown on the right side of results)
  const fullTextEl = item.querySelector('.gs_ggs a') || item.querySelector('.gs_or_ggsm a');
  const fullTextUrl = fullTextEl?.href || '';
  const fullTextType = fullTextEl?.querySelector('span.gs_ctg2')?.textContent?.trim() || '';

  // Meta info for context
  const meta = item.querySelector('.gs_a')?.textContent || '';
  const parts = meta.split(' - ');
  const authors = parts[0]?.trim() || '';
  const journalYear = parts[1]?.trim() || '';

  // Try to extract DOI from paper URL
  let doi = '';
  if (paperUrl.includes('doi.org/')) {
    doi = paperUrl.replace(/^https?:\/\/(dx\.)?doi\.org\//, '');
  }

  // Build access links
  const links = {};

  if (fullTextUrl) {
    links.fullText = fullTextUrl;
    links.fullTextType = fullTextType || (fullTextUrl.endsWith('.pdf') ? '[PDF]' : '[HTML]');
  }

  if (paperUrl) {
    links.publisher = paperUrl;
  }

  if (doi) {
    links.doi = `https://doi.org/${doi}`;
    links.scihub = `https://sci-hub.ru/${doi}`;
  } else if (paperUrl) {
    // Sci-Hub also works with direct URLs
    links.scihub = `https://sci-hub.ru/${paperUrl}`;
  }

  return {
    dataCid: cid,
    title,
    authors,
    journalYear,
    doi,
    fullTextUrl,
    fullTextType,
    paperUrl,
    links
  };
}

Step 3: Report

## Full Text Links — {title}

**Authors:** {authors} | {journalYear}

**Direct Full Text:**
{links.fullText ? "- " + links.fullTextType + " " + links.fullText : "No direct full text link available"}

**Publisher Page:**
{links.publisher ? "- " + links.publisher : "N/A"}

**DOI:**
{links.doi ? "- " + links.doi : "No DOI detected"}

**Sci-Hub:**
{links.scihub ? "- " + links.scihub : "N/A"}

Step 4: Open full text (optional)

If the user wants to read the paper immediately, use mcp__chrome-devtools__new_page to open the preferred link in this priority:

1. fullTextUrl (direct PDF/HTML, usually free)
2. DOI link (may require subscription)
3. Sci-Hub link (fallback)
4. Publisher page

Notes

• This skill uses 1-2 tool calls: evaluate_script (if already on results page) or navigate_page + evaluate_script
• Google Scholar's full-text links (.gs_ggs) are usually free/open-access PDFs
• DOI may not always be extractable from the paper URL
• Sci-Hub works with both DOI and direct URL
• Opening full text in a new tab adds 1 more tool call (new_page)