Gemini Codebase Analysis Skill

Analyze entire codebases using Gemini 2.5 Pro's massive context window (1M+ tokens). This skill collects repository files, sends them to Gemini for analysis, and returns comprehensive technical insights.

Prerequisites

•
API Key: Set your Gemini API key:
bash
```
export GEMINI_API_KEY="your-api-key"
```
•
Python SDK: Install the Google GenAI SDK:
bash
```
pip install google-genai
```

Usage

Basic Analysis with Custom Query

bash

python /Users/jms/.claude/skills/gemini-codebase-analysis/gemini_codebase_analyzer.py /path/to/repo "Your question about the codebase"

Using Analysis Templates

Available templates:

•architecture - Full architectural analysis
•security - Security vulnerability assessment
•performance - Performance analysis
•quality - Code quality assessment
•dependencies - Dependency analysis
•onboarding - Developer onboarding guide

bash

# Architecture analysis
python /Users/jms/.claude/skills/gemini-codebase-analysis/gemini_codebase_analyzer.py /path/to/repo --template architecture

# Security audit
python /Users/jms/.claude/skills/gemini-codebase-analysis/gemini_codebase_analyzer.py /path/to/repo --template security

Save Output to File

bash

python /Users/jms/.claude/skills/gemini-codebase-analysis/gemini_codebase_analyzer.py /path/to/repo --template architecture -o analysis.md

JSON Output (for programmatic use)

bash

python /Users/jms/.claude/skills/gemini-codebase-analysis/gemini_codebase_analyzer.py /path/to/repo "question" --json

Options

Option	Description
`repo_path`	Path to the repository to analyze (required)
`query`	Analysis question (required if no template)
`--template, -t`	Use predefined analysis template
`--output, -o`	Save output to file
`--model, -m`	Gemini model (default: gemini-2.5-pro)
`--max-size`	Max context size in MB (default: 50)
`--context, -c`	Additional context to include
`--json`	Output as JSON

What Gets Analyzed

The analyzer collects:

•Source code files (.py, .js, .ts, .java, .go, .rs, etc.)
•Config files (.json, .yaml, .toml, etc.)
•Documentation (.md, .rst, .txt)
•Build/DevOps files (Dockerfile, Makefile, etc.)

Automatically excluded:

•.git, node_modules, vendor, venv
•Binary files, images, videos
•Lock files, build outputs
•Files > 500KB, total context > 50MB

Example Queries

Architecture

•"How is the application structured?"
•"What design patterns are used?"
•"How does data flow through the system?"

Security

•"Are there any security vulnerabilities?"
•"How is authentication implemented?"
•"Are secrets properly managed?"

Performance

•"Where are the potential bottlenecks?"
•"Are there any N+1 query issues?"
•"How is caching implemented?"

Understanding

•"How does the [X] feature work?"
•"What happens when a user [action]?"
•"How are [X] and [Y] connected?"

Integration with Claude Code

After running the analyzer, the output can be:

•Printed directly for immediate review
•Saved to a file for reference
•Output as JSON for further processing

The analysis results are designed to be consumed by Claude Code for:

•Answering follow-up questions
•Generating implementation plans
•Identifying specific code changes needed