gemini-peer-review

To get a second opinion on code and architecture, you run a peer review on targeted files or directories. The Skill reads your code into context, constructs an AI-to-AI prompt, and calls the Gemini generateContent API using GEMINI_API_KEY to return independent analysis.

To debug logic errors and scan for security vulnerabilities, you provide the target files along with relevant error messages or prior attempts. The process generates an error-focused analysis prompt and sends it to Gemini to identify issues and common security flaws.

Yes, you need a GEMINI_API_KEY to run an independent code review. The Skill requires this API key to authenticate and call the Gemini generateContent API, which processes the constructed AI-to-AI prompt and returns the technical feedback.

Yes, you can use this approach for quick architecture questions without file context. The process supports rapid “quick question” analysis, allowing you to get design trade-off advice and recommendations directly from Gemini without reading a specific directory.

An AI-to-AI prompt for reviewing design trade-offs works by reading your target files into context, building a templated request, and writing it to .claude/artifacts/gemini-prompt.txt. Gemini then independently evaluates your design decisions and recommends safer options.