web-search · jwynia — Claude Code skill: web search with no API key
A skill that tells the agent how to use its built-in WebSearch: current info, fact checks, source links. No API key needed.
npx skills add jwynia/agent-skills --skill web-search -a claude-codeWhat web-search does
There is no search engine inside this skill. It documents how the agent should call the built-in WebSearch tool and what to do with the results: cite sources, merge findings from several results, cross-check claims, and note publication dates. So everything depends on how WebSearch is implemented on your platform. No WebSearch tool, no benefit.
The rest is query craft. Be specific and add context: not "react hooks" but "React 19 useActionState hook tutorial". For anything current, put the year in the query. A separate section covers iterative searching: when the first results are too broad, narrow the wording, try different phrasings, and pull corroborating sources.
The skill states its own limits: paywalled and login-only pages stay out of reach, freshness depends on indexing, and there is no domain filtering or relevance scoring. For those, the author points to a sibling skill, web-search-tavily, which runs on an API key. Worth checking whether the built-in search is enough before adding a second skill and a key.
What it helps with
- find information missing from the model's training data
- verify a fact and back it with a source link
- catch up on a recent release or library update
- collect several viewpoints on a contested question
- check the publication date on time-sensitive information
Example prompts for the agent
- Search the web for what shipped in React 19 and give me the source links
- Check recent coverage to see whether this library is actually abandoned
- Search a few phrasings and summarize what people write about Core Web Vitals in 2025
How to install web-search
1Skills CLI (Vercel)
npx skills add jwynia/agent-skills --skill web-search -a claude-code2Manually: put the files in place
~/.claude/skills/web-search/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/web-search/
1Skills CLI (Vercel)
npx skills add jwynia/agent-skills --skill web-search -a codex2Manually: put the files in place
.agents/skills/web-search/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add jwynia/agent-skills --skill web-search -a cursor2Manually: put the files in place
.cursor/skills/web-search/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install jwynia/agent-skills web-search2Manually: put the files in place
.github/skills/web-search/SKILL.md1Skills CLI (Vercel)
npx skills add jwynia/agent-skills --skill web-search -a gemini-cli2Platform CLI
gemini skills install https://github.com/jwynia/agent-skills --path skills/general/research/tools/web-searchManually: put the files in place
.devin/skills/web-search/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add jwynia/agent-skills --skill web-search -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about web-search
How do I install web-search in Claude?
Run npx skills add jwynia/agent-skills --skill web-search -a claude-code. The skill file then sits at ~/.claude/skills/web-search/SKILL.md.
What does web-search do?
It sets the agent's procedure for the built-in WebSearch tool: how to phrase a query, when to search again, and how to attribute sources in the answer.
Does it need an API key?
No. It relies on the platform's built-in WebSearch. If you need domain filtering and relevance scores, the author points to a separate skill, web-search-tavily, which does require a key.
Source text
SKILL.md · MIT
--- name: web-search description: "Search the web using the agent's built-in WebSearch tool. Use when you need to find current information, verify facts, or research topics. No API key required. Keywords: search, web, internet, lookup, find, research, current events, facts." license: MIT compatibility: Works with any agent that has WebSearch capability metadata: author: jwynia version: "1.0" type: utility mode: generative domain: research --- # 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
© jwynia · MIT · jwynia/agent-skills
Similar skills
All skills for Agent setup: memory, context, orchestration →web-search by other authors comes first.