Is the external code execution secure?

Yes, the skill utilizes Git worktrees for file isolation and supports Docker-based sandboxing to ensure that external tools only have access to the specific task directory.

Do I need API keys for both Codex and Gemini?

No, the skill is designed to be availability-aware. It adapts its workflow based on which tools you have configured and authenticated, falling back to standard behavior if no external tools are found.

Why use external tools instead of just Claude?

Using external tools allows for significant cost savings by routing simpler tasks to cheaper models and increases code quality by using one model to review another's work, catching errors a single model might overlook.

What happens if an external tool fails?

The skill features a robust escalation model. If an external tool fails to meet validation criteria, the task is automatically passed to the next tool in the chain or escalated to Claude for final implementation.

Cross-Model AI Orchestrator

Name: Cross-Model AI Orchestrator
Author: jpeggdev

byjpeggdev

0•

開発者ツール

Delegates coding and review tasks to external AI models like Codex and Gemini to optimize costs and improve code quality through adversarial validation.

The External Tools skill for Claude Code enables a sophisticated 'Tribunal' workflow by offloading implementation and review tasks to external AI CLI tools. By leveraging models like OpenAI Codex and Google Gemini for initial drafts and cross-model reviews, it eliminates single-model blind spots and significantly reduces operational costs. The system ensures high reliability through sandboxed execution, independent orchestrator validation of all results, and a dynamic escalation model that prioritizes the most cost-effective tools before falling back to premium models.

主な機能

01Dynamic cost-aware routing with automatic escalation logic

02Sandboxed execution using Docker or platform-level isolation

03Cross-model adversarial review to eliminate single-model blind spots

040 GitHub stars

05Independent orchestrator validation via automated testing and linting

06Multi-model delegation to OpenAI Codex and Google Gemini CLI tools

ユースケース

01Building a resilient development pipeline with multi-layered fallback and retry strategies

02Reducing API costs by offloading routine implementation tasks to cheaper secondary models

03Improving code reliability through automated peer review between different AI architectures

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jpeggdev/buildwithjpegg external-tools

For use in Claude.ai and ChatGPT

主な機能

01Dynamic cost-aware routing with automatic escalation logic

02Sandboxed execution using Docker or platform-level isolation

03Cross-model adversarial review to eliminate single-model blind spots

040 GitHub stars

05Independent orchestrator validation via automated testing and linting

06Multi-model delegation to OpenAI Codex and Google Gemini CLI tools

ユースケース

01Building a resilient development pipeline with multi-layered fallback and retry strategies

02Reducing API costs by offloading routine implementation tasks to cheaper secondary models

03Improving code reliability through automated peer review between different AI architectures