Exa Web Toolkit

What it does

• Web search: taps Exa's hybrid keyword + semantic index for topic research and current info. --category "research paper" biases results toward scholarly work, and --include-domains restricts to allowlists like arxiv.org, nature.com, or pubmed.
• URL extraction: fetches full content from webpages, articles, and academic PDFs, handling multiple URLs in a single call — better than built-in WebFetch for batches.
• Routing table: tells Claude when to search vs. extract, with a documented two-pass academic strategy.
• Citation rules: prefers author + year (e.g., [Smith et al., 2025]) and DOI links.

Who it's for

• Researchers and grad students who need peer-reviewed or preprint evidence
• Engineers who prefer standards bodies and government sources (NIH, WHO, NIST) over blog spam
• Analysts who need to pull and summarize many URLs at once

Examples

1. "Find recent papers on protein structure prediction" → runs the two-pass academic strategy with --category "research paper" and a scholarly domain allowlist.
2. "Read these three arXiv PDFs and build a methods comparison table" → batch extraction, then synthesis.
3. "Summarize the last six months of LLM evaluation benchmark news" → general web search with sourced, DOI-preferred citations.

Install the skill I found on Claude Skill Mart.
Copy the skills/exa-search folder from the GitHub repo K-Dense-AI/scientific-agent-skills into my ~/.claude/skills/exa-search/.
When it's done, tell me in one line what this skill can do.