ai-search

name	ai-search
description	Model-agnostic AI-powered web search. Use Claude's built-in WebSearch by default, or plug in your own API from OpenAI, Anthropic, Google Gemini, or Perplexity. This skill provides real-time web search with source citations for research, fact-checking, and accessing current information beyond model knowledge cutoffs.

AI Search (Model-Agnostic)

Overview

Perform AI-powered web searches using your preferred provider. This skill is model-agnostic - use Claude's built-in WebSearch by default (no API key needed), or plug in your own API from:

• Claude WebSearch (default) - Built-in, no setup required
• OpenAI (GPT-4o with web browsing)
• Anthropic (Claude API with tool use)
• Google Gemini (with Search grounding)
• Perplexity (via OpenRouter)

When to Use This Skill

Use this skill when:

• Searching for current information or recent developments
• Finding latest scientific publications and research
• Verifying facts with source citations
• Accessing information beyond the model's knowledge cutoff
• Conducting domain-specific research (biomedical, technical, clinical)
• Comparing current approaches or technologies

Do not use for:

• Simple calculations or logic problems (use directly)
• Tasks requiring code execution (use standard tools)
• Questions well within the model's training data (unless verification needed)

Quick Start

Default: Claude WebSearch (Recommended)

No setup required. Simply ask Claude to search:

Search the web for the latest developments in CRISPR gene editing in 2024-2025.

Claude will use its built-in WebSearch tool and provide results with source citations.

Advantages:

• No API key needed
• No cost to you
• Automatic source citations
• Integrated with Claude's reasoning

External Providers

For users who want to use their own API keys or specific providers.

Option 1: OpenAI

# Set API key
export OPENAI_API_KEY='sk-...'

# Search
python scripts/ai_search.py "CRISPR developments 2024" --provider openai

Option 2: Anthropic

# Set API key
export ANTHROPIC_API_KEY='sk-ant-...'

# Search
python scripts/ai_search.py "CRISPR developments 2024" --provider anthropic

Option 3: Google Gemini

# Set API key
export GEMINI_API_KEY='...'

# Search
python scripts/ai_search.py "CRISPR developments 2024" --provider gemini

Option 4: Perplexity (via OpenRouter)

# Set API key
export OPENROUTER_API_KEY='sk-or-v1-...'

# Search
python scripts/ai_search.py "CRISPR developments 2024" --provider perplexity

Provider Comparison

Provider	Setup	Cost	Strengths	Best For
Claude WebSearch	None	Free*	Integrated reasoning, citations	Default use, research
OpenAI	API key	$$	GPT-4o web browsing	General queries
Anthropic	API key	$$	Claude with tools	Complex reasoning
Gemini	API key	$	Google Search grounding	Broad web coverage
Perplexity	OpenRouter key	$$	Real-time citations	Scientific literature

*Free with Claude Code subscription

Usage Examples

Basic Search (Claude WebSearch)

Just ask naturally:

What are the latest clinical trial results for CAR-T cell therapy in 2024-2025?

External Provider Search

# OpenAI
python scripts/ai_search.py \
  "Compare PyTorch vs TensorFlow for transformers in 2024" \
  --provider openai \
  --output results.json

# Gemini with grounding
python scripts/ai_search.py \
  "Latest AlphaFold 3 accuracy improvements" \
  --provider gemini \
  --verbose

# Perplexity for scientific literature
python scripts/ai_search.py \
  "Gut microbiome Parkinson's disease research 2024" \
  --provider perplexity \
  --model sonar-pro

Crafting Effective Queries

Be Specific and Detailed

Good examples:

• "What are the latest clinical trial results for CAR-T cell therapy in treating B-cell lymphoma published in 2024?"
• "Compare the efficacy and safety profiles of mRNA vaccines versus viral vector vaccines for COVID-19"
• "Explain AlphaFold3 improvements over AlphaFold2 with specific accuracy metrics"

Bad examples:

• "Tell me about cancer treatment" (too broad)
• "CRISPR" (too vague)
• "vaccines" (lacks specificity)

Include Time Constraints

Web search excels at finding current information:

• "What papers were published in Nature Medicine in 2024 about long COVID?"
• "What are the latest developments (past 6 months) in large language model efficiency?"
• "What was announced at NeurIPS 2024 regarding AI safety?"

Specify Domain and Sources

For high-quality results, mention source preferences:

• "According to peer-reviewed publications in high-impact journals..."
• "Based on FDA-approved treatments..."
• "From clinical trial registries like clinicaltrials.gov..."

Configuration

Environment Variables

The easiest way to configure API keys is using a .env file at your project root:

Option 1: Use .env file (Recommended)

1. Copy the template: cp .env.template .env
2. Edit .env and uncomment/fill in the keys you need
3. Scripts auto-load from .env - no manual export needed!

# In your .env file:
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GEMINI_API_KEY=AIza...
OPENROUTER_API_KEY=sk-or-v1-...

Security Note: The .env file is gitignored. Never commit API keys!

Option 2: Export in shell

# Claude WebSearch - no configuration needed

# OpenAI
export OPENAI_API_KEY='sk-...'

# Anthropic
export ANTHROPIC_API_KEY='sk-ant-...'

# Google Gemini (supports both names)
export GEMINI_API_KEY='...'
# or: export GOOGLE_API_KEY='...'

# Perplexity (via OpenRouter)
export OPENROUTER_API_KEY='sk-or-v1-...'

Optional dependency for .env support:

pip install python-dotenv

Configuration File

Create ~/.ai-search.json for persistent configuration:

{
  "default_provider": "claude",
  "providers": {
    "openai": {
      "model": "gpt-4o",
      "max_tokens": 4000
    },
    "anthropic": {
      "model": "claude-sonnet-4-20250514",
      "max_tokens": 4000
    },
    "gemini": {
      "model": "gemini-2.0-flash",
      "max_tokens": 4000
    },
    "perplexity": {
      "model": "sonar-pro",
      "max_tokens": 4000
    }
  }
}

Integration with Other Skills

Literature Review

Use with literature-review skill:

1. Use AI search to find recent papers and preprints
2. Supplement PubMed searches with real-time web results
3. Verify citations and find related work

Scientific Writing

Use with scientific-writing skill:

1. Find recent references for introduction/discussion
2. Verify current state of the art
3. Check latest terminology and conventions

Citation Management

Use with citation-management skill:

1. Find DOIs for recent papers
2. Get publication details for citations
3. Verify author and venue information

Best Practices

Provider Selection

1. Start with Claude WebSearch: Free, integrated, works for most queries
2. Use external providers when: You need specific model capabilities, want to compare results, or have rate limits on Claude

Query Design

1. Be specific: Include domain, time frame, and constraints
2. Use terminology: Domain-appropriate keywords and phrases
3. Specify sources: Mention preferred publication types or journals
4. Structure questions: Clear components with explicit context

Cost Optimization

When using external providers:

1. Use Claude WebSearch for exploration: Free initial research
2. Reserve external APIs for specific needs: Targeted queries
3. Set token limits: Use --max-tokens to control costs
4. Monitor usage: Track API spending

Troubleshooting

Claude WebSearch Not Working

Claude WebSearch is built-in and should always work. If issues occur:

• Ensure you're asking a web search-appropriate question
• Try rephrasing more specifically
• Check Claude Code is up to date

API Key Errors

Error: API key not configured for provider: openai

Solution: Set the appropriate environment variable:

export OPENAI_API_KEY='sk-...'

Rate Limiting

Solution: Wait and retry, or switch to a different provider temporarily.

Invalid Model

Solution: Check available models for your provider in the configuration section.

Resources

Bundled Resources

Scripts:

• scripts/ai_search.py: Main search script with provider abstraction
• scripts/providers/: Provider implementations

References:

• references/provider_setup.md: Detailed setup for each provider
• references/query_strategies.md: Query design best practices

Assets:

• assets/.env.example: Environment variable template
• assets/config.example.json: Configuration file template

External Resources

Provider Documentation:

• OpenAI: https://platform.openai.com/docs
• Anthropic: https://docs.anthropic.com
• Google AI: https://ai.google.dev/docs
• OpenRouter: https://openrouter.ai/docs

Dependencies

For External Providers

# Install via pip
pip install openai anthropic google-generativeai litellm

# Or via requirements
pip install -r requirements.txt

Required Environment Variables

Provider	Variable	Get Key From
OpenAI	OPENAI_API_KEY	https://platform.openai.com/api-keys
Anthropic	ANTHROPIC_API_KEY	https://console.anthropic.com
Gemini	GEMINI_API_KEY	https://aistudio.google.com/apikey
Perplexity	OPENROUTER_API_KEY	https://openrouter.ai/keys

Summary

This skill provides:

1. Model-agnostic search: Use any provider you prefer
2. Zero-config default: Claude WebSearch works out of the box
3. External API support: Plug in OpenAI, Anthropic, Gemini, or Perplexity
4. Scientific focus: Optimized for research and literature search
5. Easy integration: Works with other scientific skills

Use AI-powered web search to find current information, recent research, and grounded answers with source citations.