메뉴
Expert internet research using Tavily and Brave Search. Use when the user needs current information beyond knowledge cutoff, real-time data (prices, stats, news), fact verification, documentation lookups, latest package versions, or current events.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
SQL engineering standards for writing correct, safe, cross-dialect SQL across PostgreSQL, MySQL, SQL Server, SQLite, DuckDB, and Oracle. Use when writing SQL queries, exploring schemas, running migrations, optimizing queries, transpiling between dialects, or connecting to databases via native CLIs.
MongoDB engineering standards for querying collections, building aggregation pipelines, designing document schemas, managing indexes, and administering deployments via mongosh. Use when writing MongoDB queries, exploring collections, designing schemas, optimizing performance, or managing replica sets and sharding.
Engineering planning standards for creating design docs, DRI working docs, ADRs, and weekly updates. Use when planning projects, writing design docs, creating RFCs, tracking risks, writing status updates, or reviewing engineering plans. Covers structure, operational tracking, context engineering, and scaling from startup to enterprise.
Multi-agent PR review that spawns parallel domain-specific subagents (python-expert, rust-expert, dbt-expert, dotnet-expert, kedro-expert, d2-tala-expert, ui-designer) based on changed file types, posts inline review comments on GitHub, GitLab, and Azure DevOps.
dbt engineering standards for building clean, modular, well-tested data transformations. Use when writing dbt models, configuring sources, designing project structure, writing tests, managing materializations, or preparing for dbt Fusion. Covers dbt Core 1.8–1.11, Fusion v2.0, SQL style, testing, governance, and deployment.
Langfuse observability platform standards for querying traces, managing prompts, debugging LLM applications, and accessing Langfuse data programmatically. Use when interacting with Langfuse, querying traces or sessions, managing prompts, instrumenting applications, or looking up Langfuse documentation. Covers CLI-based API access (via npx) and documentation retrieval.
| name | internet-research |
| description | Expert internet research using Tavily and Brave Search. Use when the user needs current information beyond knowledge cutoff, real-time data (prices, stats, news), fact verification, documentation lookups, latest package versions, or current events. |
This skill provides Claude Code with expert-level internet research capabilities using two complementary MCP servers: Tavily (AI-native search, extraction, crawling) and Brave Search (multimedia search, news, local).
TAVILY_API_KEY and BRAVE_API_KEY environment variables set in your shell profileSetup (if not already done):
# Add Tavily MCP to user scope
claude mcp add tavily --scope user \
-e TAVILY_API_KEY='${TAVILY_API_KEY}' \
-- npx -y tavily-mcp@latest
# Add Brave Search MCP to user scope
claude mcp add brave-search --scope user \
-e BRAVE_API_KEY='${BRAVE_API_KEY}' \
-- npx -y @brave/brave-search-mcp-server
# Verify both are configured
claude mcp list
Invoke this skill when the user needs:
These two servers are complementary tools with different jobs, not competing alternatives. Route by task type:
| Task | Tool | Why |
|---|---|---|
| Technical docs, APIs, coding questions | mcp__tavily__tavily_search | AI-native, token-efficient results |
| General/casual web search | mcp__brave-search__brave_web_search | Equal free quota (1,000/mo) |
| Image search | mcp__brave-search__brave_image_search | Brave exclusive |
| Video search | mcp__brave-search__brave_video_search | Brave exclusive |
| News search | mcp__brave-search__brave_news_search | Brave exclusive |
| Local business search | mcp__brave-search__brave_local_search | Brave exclusive |
| Extract content from URLs | mcp__tavily__tavily_extract | Tavily exclusive |
| Crawl a website | mcp__tavily__tavily_crawl | Tavily exclusive |
| Map a website's structure | mcp__tavily__tavily_map | Tavily exclusive |
| Deep multi-source research | mcp__tavily__tavily_research | Tavily exclusive |
| Either server's quota exhausted | The other one | Fallback |
mcp__tavily__tavily_searchAI-native web search. Returns pre-cleaned, structured, LLM-optimized results. Best for technical and coding queries.
Parameters:
query (required): Search termssearch_depth (optional): "basic" | "advanced" | "fast" | "ultra-fast" (default: "basic")max_results (optional): Number of results (default: 5, max: 20)time_range (optional): "day" | "week" | "month" | "year"include_domains (optional): Array of domains to includeexclude_domains (optional): Array of domains to excludecountry (optional): Boost results from specific countrymcp__tavily__tavily_extractExtract clean content from URLs in markdown or text format.
Parameters:
urls (required): Array of URLs to extract content fromextract_depth (optional): "basic" | "advanced" (use advanced for LinkedIn, protected sites)format (optional): "markdown" | "text" (default: "markdown")mcp__tavily__tavily_crawlCrawl a website with configurable depth and breadth.
Parameters:
url (required): Root URL to begin crawlingmax_depth (optional): How far from base URL (default: 1)limit (optional): Total links to process (default: 50)mcp__tavily__tavily_mapMap a website's structure, returning discovered URLs.
Parameters:
url (required): Root URL to begin mappingmcp__tavily__tavily_researchComprehensive multi-source research with AI synthesis.
Parameters:
input (required): Description of the research taskmodel (optional): "mini" | "pro" | "auto" (default: "auto")mcp__brave-search__brave_web_searchGeneral web search. Best for routine queries and preserving Tavily credits.
Parameters:
query (required): Search termscount (optional): Number of results (default: 10, max: 20)offset (optional): Pagination offsetmcp__brave-search__brave_image_searchImage search — only available via Brave.
mcp__brave-search__brave_video_searchVideo search — only available via Brave.
mcp__brave-search__brave_news_searchNews search with freshness controls — only available via Brave.
mcp__brave-search__brave_local_searchLocal business/place search — only available via Brave.
"useEffect cleanup function"Use the task-based routing table above. Key rules:
tavily_search (AI-native quality matters here)brave_web_search (preserves Tavily credits)tavily_extract (cleaner than WebFetch)tavily_research (built-in multi-source synthesis)| User Request | Tool | Effective Query |
|---|---|---|
| "How to use React Server Components" | Tavily Search | "React Server Components tutorial 2026" |
| "Find images of the new MacBook" | Brave Image | "MacBook Pro 2026" |
| "Latest tech news today" | Brave News | "technology news today" |
| "What restaurants are near Union Square" | Brave Local | "restaurants Union Square" |
| "Research WebAssembly limitations" | Tavily Research | "Current state of WebAssembly support and limitations" |
| "What's the weather in Chicago" | Brave Web | "weather Chicago today" |
| "Extract the API docs from this URL" | Tavily Extract | (pass URL) |
When presenting research results:
## Research Results
**Query:** [what you searched for]
### Key Findings
1. **[Finding 1]**
- Details...
- Source: [URL]
2. **[Finding 2]**
- Details...
- Source: [URL]
### Summary
[Concise synthesis of findings]
### Sources
- [Source 1](URL)
- [Source 2](URL)
If Tavily fails:
TAVILY_API_KEY is setmcp__brave-search__brave_web_searchIf Brave fails:
BRAVE_API_KEY is setmcp__tavily__tavily_search| Server | Free Tier | Best For |
|---|---|---|
| Tavily | 1,000 credits/mo | Technical/coding queries, extraction, crawling, research |
| Brave | 1,000 queries/mo ($5 free credits) | General web search, images, video, news, local |