Поиск научных статей через Gemini

★ 7.1 · research

gemini-search is a Claude Code skill that uses Gemini as an AI-powered literature discovery engine, surfacing research papers beyond what arXiv and Semantic Scholar keyword indexes typically return. Rather than simple keyword matching, Gemini decomposes the query into sub-problems, explores naming variants and aliases, and finds papers from multiple angles. The skill targets 15 results by default, supports filtering by publication year (default: 2022+), specific venues like NeurIPS or ICML, and a code-only mode for papers with open-source implementations. It operates via the mcp__gemini-cli__ask-gemini tool using the auto-gemini-3 model, with fallback to direct CLI, and requires Google Gemini CLI plus gemini-mcp-tool installed and authenticated. Ideal for researchers conducting broad literature reviews on emerging or cross-disciplinary topics.