jwynia/agent-skills

web-search

Search the web using the agent's built-in WebSearch tool.

Quelltext ansehen

Verwendung

  1. Installationsbefehl kopieren und das Quellpaket vor der Installation in der eigenen Agentenumgebung prüfen.
  2. Benötigte Werkzeuge, Zugangsdaten und Hosts im Original prüfen.
  3. Eingabe und gewünschtes Ergebnis beschreiben und die Ausgabe mit den Quellen abgleichen.
npx skills add https://github.com/jwynia/agent-skills --skill web-search

Vor dem Start

  • Die Installation umfasst weder Modellabo noch API-Kontingent oder Dateizugriffsrechte.
  • Sterne und Installationen zeigen Aktivität, keine unabhängige Qualitäts- oder Sicherheitsprüfung.
Originales Skill-Dokument

Aus dem Quell-Repository gerendert; Überschriften, Beispiele, Code, Tabellen, Links und Bilder bleiben erhalten.

Web Search

Search the web using the agent's built-in WebSearch capability. No external API keys required.

When to Use This Skill

Use this skill when:

  • You need to find current information not in your training data
  • The user asks about recent events, news, or updates
  • You need to verify facts or find sources
  • Research requires real-time web data

Do NOT use this skill when:

  • Information is already in your knowledge base
  • You're working with local files or code
  • You need advanced filtering (use web-search-tavily instead)
  • A more specific research skill applies

How to Search

Use the agent's built-in WebSearch tool directly. The tool accepts a query string and returns relevant web results.

Basic Search

Simply invoke WebSearch with your query:

Query: "React 19 new features"

Effective Query Strategies

Be specific and include context:

  • Bad: "react hooks"
  • Good: "React 19 useActionState hook tutorial"

Include the year for current information:

  • Bad: "best TypeScript practices"
  • Good: "TypeScript best practices 2025"

Use domain-specific terms:

  • Bad: "how to make website fast"
  • Good: "web performance optimization Core Web Vitals"

When to Search Multiple Times

Search iteratively when:

  1. Initial results are too broad → Refine with more specific terms
  2. Looking for multiple perspectives → Search different phrasings
  3. Verifying facts → Search for corroborating sources
  4. Deep research → Start broad, then drill into specifics

Output Handling

After receiving search results:

  1. Cite sources - Always include URLs when sharing information
  2. Synthesize - Combine information from multiple results
  3. Verify - Cross-reference claims across sources
  4. Date-check - Note publication dates for time-sensitive information

Source Attribution Format

When sharing information from search results:

According to [Source Name](URL), ...

Sources:
- [Title 1](url1)
- [Title 2](url2)

Limitations

  • Results depend on the agent's WebSearch implementation
  • Cannot access paywalled or login-required content
  • May not have the most recent information (depends on indexing)
  • No domain filtering or relevance scoring (use web-search-tavily for these features)

Related Skills

  • web-search-tavily - Advanced search with API key, domain filtering, and relevance scores
  • research-workflow - Structured research with planning and synthesis
  • fact-check - Verify specific claims against sources
  • claim-investigation - Investigate viral claims and social media content

Verwandte Aufgaben

aus demselben Repository

Weitere Skills

Alle Skills