gemini

Overview

Gemini is a skill from the softaworks/agent-toolkit that enables running Gemini CLI from within Claude Code for code review, plan review, and big context processing requiring over 200K tokens. It uses Gemini 3 Pro by default for state-of-the-art reasoning and coding performance, with multiple model options and approval modes for different use cases.

Key Features

• Large Context Processing - Designed for tasks requiring over 200K tokens of context, such as entire codebases, documentation sets, or multi-file analysis
• Multiple Model Options - Supports Gemini 3 Pro Preview (flagship), Gemini 3 Flash (sub-second latency), and legacy 2.5 Pro, 2.5 Flash, and 2.5 Flash Lite options
• Background Safety Warnings - Critical guidance to never use --approval-mode default in non-interactive shells as it will hang indefinitely, with --approval-mode yolo required for automated tasks
• Hung Process Detection - Built-in diagnosis for symptoms of hung Gemini processes (20+ minutes with 0% CPU, sleeping state) with kill commands
• Flexible Approval Modes - Three approval modes: default (interactive only), auto_edit (auto-approve edits), and yolo (auto-approve all) with sandbox isolation option

Who is this for?

This is for Claude Code users who need to leverage Gemini's large context window for comprehensive code reviews, architectural plan analysis, or processing entire codebases. It is ideal for developers who want to combine Claude Code's workflow with Gemini's 200K+ token context processing for tasks that exceed Claude's context limits.