Can I choose specific AI models for the review?

Yes, the skill allows you to pin specific model versions, such as gpt-5.3-codex or gemini-3.1-pro-preview, to ensure high-fidelity analysis.

Why should I use external agents instead of just Claude?

Different AI models have different strengths and biases; getting a 'second opinion' helps catch edge cases, security vulnerabilities, or architectural smells that a single model might overlook.

What tools are required to use this skill?

You must have the Codex CLI (@openai/codex) and Gemini CLI (gemini-cli) installed and authenticated on your local machine.

Is it possible to run these reviews without modifying my code?

Absolutely. The skill supports a 'read-only' sandbox mode for Codex and non-interactive prompts for Gemini to ensure analysis remains strictly observational.

How does this help with Git workflows?

The skill includes native commands to review uncommitted changes, specific commit SHAs, or compare your current branch against the main branch automatically.

External AI Agents

Name: External AI Agents
Author: kylesnowschwartz

bykylesnowschwartz

•

Collaboration Tools

Integrates OpenAI Codex and Google Gemini CLIs to provide multi-model code reviews and architectural analysis within Claude Code.

The External AI Agents skill empowers Claude Code to orchestrate secondary AI perspectives for more robust development workflows. By leveraging specialized CLIs like OpenAI Codex and Google Gemini, this skill provides 'second opinions' on complex diffs, identifies architectural flaws, and conducts security audits from varied model perspectives. It supports Git-aware reviews, sandboxed execution, and specific model pinning, ensuring that critical code changes are validated by the industry's most advanced reasoning models beyond the primary Claude instance.

Key Features

01Advanced model pinning for specific versions like gpt-5.3-codex and gemini-3.1.

02Native integration with OpenAI Codex for SWE-optimized code reviews.

03Seamless Google Gemini CLI orchestration for alternative architectural analysis.

04Git-aware diff analysis for uncommitted changes and branch comparisons.

0591 GitHub stars

06Configurable execution modes including read-only sandboxes and full-auto mode.

Use Cases

01Requesting an independent second opinion on a complex Pull Request before merging.

02Performing automated security audits and race-condition checks using multiple AI perspectives.

03Validating architectural design decisions across different LLM families (OpenAI vs Google).

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add kylesnowschwartz/simpleclaude external-agents

For use in Claude.ai and ChatGPT

Download Skill