vendor-risk-kyc-screening

作者: tavily-ai

Screen vendors, merchants, suppliers, counterparties, companies, executives, and related entities for vendor risk, KYC, adverse media, sanctions, regulatory…

npx skills add https://github.com/tavily-ai/use-case-skills --skill vendor-risk-kyc-screening

Vendor Risk KYC Screening

Workflow

Use search and extract for source-grounded vendor, merchant, supplier, counterparty, and KYC research; use map or crawl only for known official directories or registries. Treat this as research support, not a final compliance determination. Execution mechanics should come from companion endpoint skills.

Treat the guidance below as base guidance; adapt it to the user's request when appropriate.

  • Identify the entity, aliases, parent/subsidiaries, executives, jurisdictions, products, and risk categories.
  • Break the screen into short subqueries under 400 characters for each alias and risk type: sanctions, enforcement, litigation, regulatory warning, recall, adverse media, cybersecurity incident, supplier risk, and jurisdiction.
  • Use exact-match style queries for legal names, people, product names, and phrases that must appear verbatim.
  • Search first, then filter by domain trust and source type before extracting. Prioritize official regulators, sanctions lists, court records, recall databases, company disclosures, and credible news.
  • Extract selected sources that can support or rule out specific findings.
  • Use site navigation only when a known regulator, registry, or company site needs URL discovery.
  • Collect multiple pages only for scoped official directories, recalls/advisories, policy pages, or supplier/product pages.

Research Budget

  • Start with a small focused search set covering entity aliases, sanctions/regulatory risk, adverse media, and jurisdiction-specific risk.
  • Extract only the strongest official or credible sources before drafting.
  • Add more searches only for named gaps, such as missing alias coverage, missing jurisdiction coverage, or unresolved high-risk findings.
  • Do not use map unless a known regulator, registry, sanctions portal, or company site has specific buried pages to locate.
  • Do not use crawl unless the user asks for scoped collection from an official directory, registry, recalls page, or supplier/product section.

Capability Guidance

  • Use search for most vendor screening, KYC, adverse media, sanctions, enforcement, litigation, recall, and supplier-risk discovery.
  • Use extract on selected official records, regulator pages, sanctions pages, court records, recall databases, company disclosures, and credible news.
  • Use map for official registries, regulator sites, sanctions portals, or company domains with hard-to-find pages.
  • Use crawl for directories or official record sets only after narrowing the paths and risk categories.
  • Use research for full due diligence reports, then verify high-impact claims against original sources.

Query And Source Guidance

  • Query legal names, trade names, aliases, parent/subsidiary names, executives, product names, and jurisdiction terms separately.
  • Combine entity terms with specific risk concepts: sanctions, enforcement action, consent order, lawsuit, fraud, recall, breach, warning letter, import alert, debarment, bankruptcy, and adverse media.
  • Prioritize regulators, sanctions databases, court systems, recall databases, official registries, company disclosures, and credible news.
  • Treat unsourced aggregators, copied press releases, and name-match-only hits as low confidence until corroborated.
  • Report failed or inaccessible sources when they affect official registries, sanctions pages, or high-risk findings.

Output Template

Use this markdown structure and avoid implying clearance:

# Vendor / KYC Risk Brief: <entity>

## Scope
- Entity and aliases checked:
- Jurisdictions:
- Risk categories:
- Source coverage limits:

## Findings
### High
- Finding:
- Evidence:
- Confidence:

### Medium
- Finding:
- Evidence:
- Confidence:

### Low / Watch
- Finding:
- Evidence:
- Confidence:

## No Relevant Findings In Searched Sources
- <risk category or alias checked>

## Recommended Next Checks
- <next check>

## Sources
- [Source title](URL) - <what it supports>

Say "no relevant findings found in searched sources" instead of "cleared" unless the user supplied authoritative internal checks.

來自 tavily-ai 的更多技能

research
tavily-ai
針對任何主題進行全面研究,自動收集來源、分析並提供引用。執行多來源網路研究並附上明確引用,適合比較、時事、市場分析及詳細報告。提供三種模型選項:mini 針對單一主題的目標研究(約30秒)、pro 進行全面的多角度分析(約60-120秒),以及 auto 透過 API 驅動的複雜度偵測。透過 Tavily MCP 伺服器以 OAuth 進行驗證,並在...上自動執行基於瀏覽器的登入。
official
search
tavily-ai
使用LLM優化結果的網路搜尋,具備相關性評分與靈活篩選功能。支援四種搜尋深度模式(極速、快速、基本、進階),可配置延遲與相關性權衡。包含網域篩選、時間範圍限制、日期區間、國家加權及原始內容提取。回傳結果包含標題、網址、內容摘要與相關性評分;可選圖片結果與網站圖示。透過Tavily MCP伺服器或API金鑰配置自動進行OAuth驗證;...
official
tavily-best-practices
tavily-ai
專為LLM設計的網路搜尋API,具備即時資料存取、內容擷取、網站爬取及AI驅動研究功能。五大核心方法:search()用於搜尋網頁結果、extract()用於擷取URL內容、crawl()用於全站擷取、map()用於URL探索,以及research()用於端到端AI綜合分析。支援Python與JavaScript SDK,提供非同步客戶端以進行平行查詢,並可設定搜尋深度(極速/快速/基本/進階)。Crawl方法接受語意指令,以聚焦於特定內容的擷取...
official
tavily-cli
tavily-ai
透過 Tavily CLI 進行網路搜尋、內容提取、網站爬取與深度研究。五種指令模式涵蓋搜尋、提取、URL 發現、批量爬取及附引用來源的多來源研究。所有指令皆支援 JSON 輸出與檔案儲存,適用於結構化、代理式工作流程。升級模式引導您從簡單搜尋,逐步進展至提取、映射、爬取,乃至依需求進行的全面研究。需安裝 tavily-cli 並透過 tvly login 進行 API 金鑰驗證。
official
tavily-crawl
tavily-ai
多頁面網站爬蟲,具備語意過濾與Markdown匯出功能。可透過深度與廣度控制爬取整個網站區塊;依路徑正則表達式、網域或自然語言指令進行過濾,以聚焦結果。透過--output-dir將每個頁面儲存為本機Markdown檔案,或回傳結構化JSON供代理處理。使用語意指令搭配區塊提取,避免將結果餵入LLM時發生上下文膨脹;採用全頁提取進行離線文件下載。支援...
official
tavily-dynamic-search
tavily-ai
搜尋網路、篩選結果並擷取內容,讓原始搜尋資料絕不進入你的上下文視窗。只有你精心整理的 print() 輸出會回傳。
official
tavily-extract
tavily-ai
從最多20個URL中提取乾淨的Markdown或純文字,支援JavaScript渲染與查詢聚焦區塊切割。可處理JavaScript渲染頁面,並提供可設定的提取深度(基本模式適用於簡單頁面,進階模式適用於動態SPA與表格)。支援查詢聚焦提取,僅回傳相關內容區塊而非完整頁面。預設回傳經LLM最佳化的Markdown格式,亦可選擇純文字格式與結構化JSON輸出。單次呼叫可處理最多20個URL;...
official
tavily-map
tavily-ai
快速發現網站上的URL,無需提取內容,非常適合在大型網站上尋找特定頁面。返回域上所有URL的結構化列表,具有可配置的深度和廣度、正則表達式路徑過濾以及用於語義過濾的自然語言指令。支援深度控制(1–5層)、每頁廣度限制、外部鏈接包含/排除,以及通過正則表達式模式進行域過濾。設計為工作流程中的第一步:先映射找到正確頁面,再使用提取或...
official