Wos Search

WoS Search — 全流程技能

一个技能覆盖 Web of Science 站内全部动作。检索在本文件；其余场景按路由表进入对应子流程文档。

场景路由

Important: Browser Prerequisite

The browser must be on any webofscience.com page (logged in). The skill calls the WoS internal API directly via fetch, so no page navigation is needed.

Language Guidance

WoS databases (especially SCI/SSCI) primarily index English-language literature. If the user provides Chinese keywords, translate to English (e.g., "价值共创" → "value co-creation"). Inform the user about the translation.

Parameter Reference

Database (product)

Edition filtering (Core Collection only)

Multiple editions can be combined: ["WOS.SCI", "WOS.SSCI"]

Sort options

Field mapping (for query rows)

Steps

Step 1: Build API Request Body

jsonCopy code
{
  "product": "WOSCC",
  "searchMode": "general",
  "viewType": "search",
  "serviceMode": "summary",
  "search": {
    "mode": "general",
    "database": "WOSCC",
    "query": [
      {"rowField": "TS", "rowText": "USER_QUERY"}
    ],
    "editions": ["WOS.SSCI"]
  },
  "retrieve": {
    "count": 10,
    "history": true,
    "jcr": true,
    "sort": "times-cited-descending",
    "analyzes": [],
    "locale": "en"
  },
  "eventMode": null
}

• query: array of {rowField, rowText} objects. For multiple conditions, add rowBoolean (AND/OR/NOT) to subsequent rows.
• editions: optional, omit for all editions.
• count: number of records to retrieve (default 10, max 50).
• sort: default relevance.

Step 2: Execute API Call via evaluate_script

This is the only tool call needed — 1 call total.

If the browser was previously on a non-WoS page (e.g., after following a publisher link), SID will be lost. In that case, first navigate back to any WoS page (navigate_page to https://www.webofscience.com/wos/woscc/basic-search) to re-establish the session, then run the API call. This adds 1 extra tool call (2 total).

Alternatively, use the URL-based fallback (Step 2B below) which always works regardless of SID state.

javascriptCopy code
async () => {
  // Extract SID from network history
  const sid = performance.getEntriesByType('resource')
    .filter(r => r.name.includes('SID='))
    .map(r => r.name.match(/SID=([^&]+)/)?.[1])
    .filter(Boolean)[0] || '';

  if (!sid) return { status: 'no_session', message: 'SID lost (likely navigated to external site). Use Step 2B or navigate to any WoS page first.' };

  const response = await fetch(`/api/wosnx/core/runQuerySearch?SID=${sid}`, {
    method: 'POST',
    headers: { 'Content-Type': 'text/plain;charset=UTF-8', 'Accept': 'application/x-ndjson' },
    body: JSON.stringify({
      "product": "{PRODUCT}",
      "searchMode": "general",
      "viewType": "search",
      "serviceMode": "summary",
      "search": {
        "mode": "general",
        "database": "{PRODUCT}",
        "query": [{QUERY_ROWS}],
        "editions": [{EDITIONS}]
      },
      "retrieve": {
        "count": {COUNT},
        "history": true,
        "jcr": true,
        "sort": "{SORT}",
        "analyzes": [],
        "locale": "en"
      },
      "eventMode": null
    })
  });

  const text = await response.text();
  const lines = text.trim().split('\n').map(line => {
    try { return JSON.parse(line); } catch(e) { return null; }
  }).filter(Boolean);

  const searchInfo = lines.find(l => l.key === 'searchInfo')?.payload;
  const recordsData = lines.find(l => l.key === 'records')?.payload;

  let records = [];
  if (recordsData) {
    records = Object.entries(recordsData).map(([idx, rec]) => ({
      idx: parseInt(idx),
      wosId: rec.colluid,
      title: rec.titles?.item?.en?.[0]?.title || '',
      authors: rec.names?.author?.en?.filter(Boolean).map(a => a.wos_standard).join('; ') || '',
      source: rec.titles?.source?.en?.[0]?.title || '',
      year: rec.pub_info?.pubyear || '',
      vol: rec.pub_info?.vol || '',
      issue: rec.pub_info?.issue || '',
      pages: rec.pub_info?.page_no || '',
      doi: rec.doi || '',
      citations: rec.citation_related?.counts?.WOSCC || 0,
      citationsAll: rec.citation_related?.counts?.ALLDB || 0,
      refCount: rec.ref_count || 0,
      abstract: rec.abstract?.basic?.en?.abstract?.replace(/<[^>]*>/g, '')?.substring(0, 300) || '',
      docType: rec.doctypes?.[0] || '',
      oa: rec.oa || false
    }));
  }

  return {
    status: 'ok',
    totalResults: searchInfo?.RecordsFound || 0,
    recordsSearched: searchInfo?.RecordsSearched || 0,
    queryId: searchInfo?.QueryID || '',
    records
  };
}

Step 3: Present Results

Display as a table:

Copy code
Found **{totalResults}** results in {database} ({edition}).
Sorted by: {sort}.

| # | Title | Authors | Source | Year | Cited | WoS ID |
|---|-------|---------|--------|------|-------|--------|
| 1 | {title} | {authors} | {source} | {year} | {citations} | {wosId} |
| ... |

Offer next actions:

• "Use /wos-paper-detail {WoS ID} for detailed information"
• "Export results: collect entries, then use the scansci-pdf skill (scansci_pdf_citation / scansci_pdf_zotero_push)"
• To see more results, re-run with higher count or use /wos-navigate-pages

Step 2B: URL-based Fallback (when SID is lost)

If API returns no_session (e.g., after navigating to an external publisher site), fall back to URL navigation:

Copy code
navigate_page({
  url: "https://www.webofscience.com/wos/{db}/general-summary?queryJson={ENCODED_QUERY_JSON}",
  initScript: "Object.defineProperty(navigator, 'webdriver', {get: () => undefined})"
})

Then extract results via evaluate_script with DOM selectors (see wos-parse-results Mode B), or re-attempt the API call (SID will be re-established after the navigation).

When to use: After the browser visited an external site (publisher, DOI link, etc.) and performance.getEntriesByType('resource') no longer contains WoS SID entries.

Notes

• 1 tool call (API) or 2 tool calls (navigate + API/DOM) if SID is lost
• API returns NDJSON (newline-delimited JSON); parse line by line
• SID is extracted from browser's performance resource entries
• SID loss: Navigating to external sites clears performance entries. Recover by navigating to any WoS page first, or use URL-based fallback (Step 2B)
• editions field uses format WOS.SCI, WOS.SSCI, WOS.CPCI-S, etc.
• For Chinese keywords in SCI/SSCI, translate to English first
• Max 50 records per request; use retrieve.count to control
• history: true saves the search to WoS session history