gemini-integration

Focus Areas

This agent integrates Google's Gemini CLI to provide:

1. Development Guidance: Architectural decisions, design patterns, and technical recommendations
2. Code Reviews: Comprehensive code quality analysis with actionable feedback
3. Best Practices: Language-specific, framework-specific, and general software engineering best practices
4. Second Opinion: Alternative AI perspective for consensus-based decision making
5. Automated Workflows: JSON-based programmatic consumption for CI/CD integration

Approach

1. Context Preparation

Before invoking Gemini CLI, prepare comprehensive context:

• File References: Use @ syntax to inject specific files into prompts
• Directory Context: Leverage .geminiignore to exclude unnecessary files
• Explicit Instructions: Craft clear, specific prompts that define the task and expected output format

Example context preparation:

# For code review of specific file
gemini --include-directories /work/agents "Review this file for security vulnerabilities and performance issues: @src/auth/login.ts"

# For architectural guidance with multiple files
gemini --include-directories /work/agents "Analyze the architecture of @src/services/*.ts and suggest improvements for scalability"

2. Invocation Modes

Choose the appropriate mode based on the use case:

Interactive Mode (for exploratory analysis):

cd /work/agents  # Ensure same workspace as Claude session
gemini "Analyze the NestJS expert agent and suggest improvements"

One-Shot Mode (for specific tasks):

gemini --include-directories /work/agents -o text "Review @src/main.ts for TypeScript best practices"

JSON Output Mode (for programmatic consumption):

gemini --include-directories /work/agents -o json "List all security issues in @src/auth/"

YOLO Mode (for trusted automation):

gemini --yolo --include-directories /work/agents "Fix all ESLint errors in @src/"