librarian_skill

Overview

This skill is a web research agent focused on documentation lookup, web searches, and open-source code discovery. It uses specialized connectors for official docs, general web search, code search via grep.app, and browser-based page scraping. It’s strictly read-only and designed to return structured evidence, links, and concise answers. Use it when you need authoritative references, code examples, or up-to-date technical signals from the web.

How this skill works

I inspect official documentation sources, run web searches for recent technical articles and issues, perform code searches on grep.app, and fetch pages that require JavaScript rendering via automated browsing. Requests are classified into types (concepts, external source analysis, diagnostics, or broad research) and routed to the appropriate tools in parallel when needed. Results are returned as structured analysis, linked evidence, a concise answer, and any uncertainties.

When to use it

• Lookup official API behavior, configuration, or best-practice guidance
• Investigate recent language/framework changes or RFCs
• Search for open-source implementations or code snippets via grep.app
• Diagnose error messages by locating related issues and published fixes
• Gather evidence and canonical links for technical decision-making

Best practices

• Provide the exact error messages, versions, and minimal context to speed diagnosis
• Specify desired outputs (e.g., official docs, code examples, links) in the prompt
• For deep codebase work or local repo search, use a local code-exploration agent instead
• Allow up to a minute for complex pages that require JS rendering
• Prefer focused questions to get direct answers and evidence links

Example use cases

• "Show React useEffect best practices with official docs and examples"
• "What’s new in TypeScript 5.5 and where is it documented?"
• "Find TanStack Query useQuery implementations across open-source projects"
• "Investigate a Zod validation error and link to related GitHub issues"
• "Deeply read and summarize a long technical blog post that requires JS rendering"