菜单
This skill should be used when the user asks to search for current information, find the latest version of a package, check for security advisories, or look up recent documentation. Relevant when the user says "what's the latest version of X," "are there any CVEs for Y," or "search for the current docs on Z."
This skill should be used when the user asks to review code changes, review a PR, get a second opinion on a diff, cross-check modifications, or perform a security review using Gemini. Relevant when the user says "review this PR with Gemini," "check my diff for bugs," or "get a second opinion on these changes."
This skill should be used when the user asks to analyze a large codebase, audit a full repository, review architecture across many files, map dependencies, or plan a migration involving more code than fits in Claude's context. Relevant when the user says "analyze this entire repo," "audit the architecture," or "this codebase is too big to review."
This skill should be used when the user asks to analyze a video, process images, transcribe audio, read or summarize a PDF, extract text from a screenshot, convert a diagram to code, or perform any visual analysis. Relevant when the user says "transcribe this audio file," "what's in this video," or "turn this diagram into code."
| name | gemini-research |
| description | This skill should be used when the user asks to search for current information, find the latest version of a package, check for security advisories, or look up recent documentation. Relevant when the user says "what's the latest version of X," "are there any CVEs for Y," or "search for the current docs on Z." |
Invoke using /gemini-search or mcp__gemini__gemini_execute with extensions: ['google_search'].
Use Gemini with Google Search grounding when you need:
If the question is about stable, well-established knowledge that hasn't changed recently, Claude's training data is likely sufficient. Use Gemini search when recency matters.
['google_search'] — this is what enables groundingflash for speed. Use pro when the query requires
synthesizing information from multiple sources or reasoning about conflicting
information.flash lookups. Set 2400000 (40 minutes) for deep pro research with multi-source synthesis.Specific queries get better results:
Include version numbers, date ranges, and specific aspects to search for.