Skills web-search
🔍

web-search

Safe ⚙️ External commands🌐 Network access

Search the Web with AI-Powered Tools

Also available from: inference-sh-9,inference-sh,inferen-sh,Cain96

Enhance your AI agents with real-time web search and content extraction capabilities. This skill provides access to Tavily and Exa APIs through inference.sh CLI for research, fact-checking, and RAG pipelines.

Supports: Claude Codex Code(CC)
🥉 74 Bronze
1

Download the skill ZIP

2

Upload in Claude

Go to Settings → Capabilities → Skills → Upload skill

3

Toggle on and start using

Test it

Using "web-search". Search for the latest developments in quantum computing

Expected outcome:

Found 5 recent sources. Key developments include IBM's 1000-qubit processor breakthrough, Google's quantum error correction improvements, and new quantum cryptography standards. Sources: Nature, MIT Technology Review, ArXiv.

Using "web-search". Extract and summarize the content from https://example.com/research-paper

Expected outcome:

Extracted 4,200 words from the research paper. Summary: The paper presents a new approach to transformer architecture that reduces computational complexity by 40% while maintaining accuracy benchmarks.

Using "web-search". What is the current population of Tokyo according to official sources?

Expected outcome:

According to the Tokyo Metropolitan Government statistics (2024), the population of Tokyo is approximately 14.1 million people in the prefecture, with 9.8 million in the 23 special wards.

Security Audit

Safe
v1 • 4/21/2026

All 38 static analysis findings are false positives from documentation code blocks and URL references. The skill legitimately uses Bash tool with inference.sh CLI for web search capabilities. No actual command injection, credential exfiltration, or weak cryptography detected.

1
Files scanned
151
Lines analyzed
5
findings
1
Total audits
Low Risk Issues (3)
Documentation Code Blocks Trigger Pattern Detectors
Backtick patterns in bash code examples (lines 15-143) triggered external_commands detection. These are documentation examples in fenced code blocks, not actual Ruby/shell backtick execution. The skill uses Bash(infsh *) tool which is properly sandboxed.
Documentation URLs Trigger Network Detection
URLs in documentation (inference.sh links, example URLs in code blocks) triggered hardcoded URL detection. These are legitimate documentation references and example parameters, not hardcoded malicious endpoints.
False Positive Weak Cryptography Detection
Static analyzer reported weak crypto (MD5) at lines 3, 29, 36, 148. Manual review confirms no MD5 usage in skill. These appear to be pattern-matching errors on line numbers or other non-crypto content.

Risk Factors

⚙️ External commands (1)
🌐 Network access (1)
Audited by: claude

Quality Score

38
Architecture
100
Maintainability
87
Content
50
Community
99
Security
91
Spec Compliance

What You Can Build

Research Agent with Source Citations

Build an AI agent that searches the web for current information on any topic, extracts relevant content, and provides answers with proper source citations. Useful for students, researchers, and analysts who need accurate, up-to-date information.

Fact-Checking Assistant

Verify claims and statements by searching for authoritative sources. The skill can extract content from multiple URLs and cross-reference information to determine accuracy. Ideal for journalists, editors, and content moderators.

RAG Pipeline Data Enrichment

Enhance retrieval-augmented generation systems by fetching fresh web content to supplement static knowledge bases. Search for recent developments, extract relevant articles, and feed them into LLM prompts for more accurate responses.

Try These Prompts

Basic Web Search
Search the web for information about [your topic]. Use Tavily Search Assistant to find relevant, recent sources.
Multi-Source Research
Search for recent developments in [topic/industry]. Extract content from the top 5 most relevant URLs and summarize the key trends.
Fact-Checking Query
Verify this claim: [specific claim or statement]. Search for authoritative sources that confirm or refute this information.
RAG-Enhanced Analysis
First search for latest information about [topic], then use those results to answer: [specific question]. Include sources in your response.

Best Practices

  • Always verify search results by checking multiple sources, especially for factual claims
  • Use specific, targeted queries rather than broad questions to get more relevant results
  • Chain search operations with LLM analysis to synthesize information from multiple sources
  • Respect rate limits and implement caching for frequently searched queries

Avoid

  • Do not use web search as the primary source for sensitive or critical decisions without human verification
  • Avoid searching for very broad topics without specific constraints, as this may return irrelevant results
  • Do not extract content from websites that prohibit scraping in their robots.txt or terms of service
  • Never rely solely on AI-generated answers without reviewing the cited sources for accuracy

Frequently Asked Questions

What API keys do I need to use this skill?
You need an inference.sh account and API key. The skill uses inference.sh CLI which handles authentication with Tavily and Exa APIs. Run 'infsh login' to configure your credentials.
Is there a limit on how many searches I can perform?
Rate limits depend on your inference.sh account tier. Free tiers typically have lower limits. Check your account dashboard for specific quotas and consider upgrading for higher volume usage.
Can I use this skill for commercial applications?
Yes, this skill is MIT-licensed. However, ensure your usage complies with inference.sh terms of service and the underlying Tavily/Exa API agreements. Commercial use may require appropriate API tiers.
What is the difference between Tavily and Exa?
Tavily excels at AI-powered search with direct answers and source citations. Exa specializes in semantic search with highly relevant results and direct factual answers. Both offer content extraction. Use Tavily for research with answers, Exa for precision search.
How do I combine search results with LLM analysis?
Use the workflow examples in the skill documentation. First run a search app and save results to a file, then pass that file as input to an LLM app. This enables RAG pipelines where web content enriches LLM context.
Does this skill work offline?
No, this skill requires internet connectivity to function. All search and extraction operations make real-time API calls to inference.sh, which then queries Tavily/Exa services. Ensure your network connection is stable before use.

Developer Details

File structure

📄 SKILL.md