gemini-collab

Coordinate Claude Code and Gemini CLI through a four-phase design, coding, review, and apply workflow.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/dbaek-star/claude-gemini-skills --skill gemini-collab
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: gemini-collab
Source: https://github.com/dbaek-star/claude-gemini-skills/tree/main/skills/gemini-collab
Command: npx skills add https://github.com/dbaek-star/claude-gemini-skills --skill gemini-collab

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill enables Claude Code and Gemini CLI to collaborate on code writing, streamlining multi-AI development workflows and reducing token usage by leveraging Subagent mode.

Core Features & Use Cases

  • Deterministic 4-phase collaboration: design → code → review → verify with automatic progress updates.
  • Subagent mode for token optimization: runs collaboration steps inside a subagent and returns only summarized results to the main context.
  • Two-stage verification: Claude reviews Gemini feedback and decides whether to adopt changes, improving output quality.
  • Audit-friendly logs & artifacts: saves run traces under .gemini/ to support reproducibility and auditing.
  • Use case examples: code writing, design discussions, and collaborative coding between Claude Code and Gemini CLI.

Quick Start

Start a collaborative coding session by invoking the gemini-collab skill in your Claude environment. To begin, request: "Initialize Gemini collaboration for coding tasks with subagent optimization."

Frequently Asked Questions about gemini-collab

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I use Claude Code and Gemini CLI together for pair programming?

Pair programming with Claude Code and Gemini CLI is enabled by a deterministic four-phase collaboration process: design, code, review, and verify, producing shared AI-generated software artifacts within a controlled workflow.

Can I reduce token usage when running AI-assisted code writing tasks?

You can reduce token usage during AI-assisted code writing by executing collaboration steps inside a subagent, which processes design and coding tasks internally and returns only summarized results to the main context.

How does the code review process work when collaborating between Claude Code and Gemini CLI?

The code review process uses a two-stage verification where Claude reviews Gemini feedback and decides whether to adopt changes, improving output quality before applying the final code modifications.

Do I need subagent mode to run collaborative coding sessions with multiple AI models?

Subagent mode is not strictly required but is highly recommended for collaborative coding sessions to minimize main-context token usage and maintain audit-friendly logs and artifacts under the .gemini/ directory.

What is the best way to audit and reproduce AI-generated software artifacts from a coding session?

To audit and reproduce AI-generated software artifacts, access the run traces saved under the .gemini/ directory, which support reproducibility by logging the deterministic phase-based collaboration steps and model fallbacks.

What happens if one AI model fails during a multi-AI code writing workflow?

If one AI model fails during a multi-AI code writing workflow, the deterministic phase-based process utilizes a model fallback mechanism to ensure the design, coding, review, and apply phases continue without interruption.