What is PROXY_MODE in Claude Code?

PROXY_MODE is a directive used at the start of a prompt that tells a Claude agent to delegate its current task to an external AI model via the claudish tool.

Can I check if a specific agent supports proxy mode?

Yes, you can run a grep command in the terminal to look for 'proxy_mode_support' within the agent's markdown file in the plugins directory.

Which agents support the PROXY_MODE directive?

Currently, 18 agents across the agentdev, frontend, and seo plugins support PROXY_MODE, including reviewers, architects, and developers. General-purpose agents typically do not support this feature.

Why did my agent respond as Claude instead of the external model?

This usually happens if you use an agent that doesn't have proxy support (like general-purpose) or if the PROXY_MODE directive was not the very first line of your prompt.

How do I format a proxy request correctly?

The directive must be the first line: 'PROXY_MODE: {model_id}' followed by a blank line and then your actual task. For example: 'PROXY_MODE: x-ai/grok-code-fast-1'.

Claude Code PROXY_MODE Orchestrator

Name: Claude Code PROXY_MODE Orchestrator
Author: involvex

byinvolvex

0•

開発者ツール

Provides a comprehensive reference and implementation guide for delegating tasks to external AI models using the PROXY_MODE directive.

The proxy-mode-reference skill is an essential technical guide for developers looking to implement multi-model workflows within the Claude environment. It explains how to use the PROXY_MODE directive to delegate specific tasks—such as code reviews, architecture validation, or SEO analysis—to external models like Grok via the claudish utility. By detailing supported agents across specialized plugins, proper prompt formatting, and common troubleshooting steps, this skill enables seamless orchestration between Claude and other AI providers for more robust and diverse output.

主な機能

01Standardized PROXY_MODE directive syntax for external model delegation

02Detailed inventory of 18+ supported agents across dev, frontend, and SEO domains

03Verification commands to check for proxy support in specific agent files

040 GitHub stars

05Best practice patterns for background task execution with external models

06Troubleshooting guide for common syntax errors and agent compatibility issues

ユースケース

01Running multi-model code reviews to get diverse perspectives on implementation quality

02Delegating specialized architecture tasks to external models like Grok within a Claude-managed workflow

03Debugging complex orchestration issues when agents fail to trigger external model calls

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add involvex/involvex-claude-marketplace proxy-mode-reference

For use in Claude.ai and ChatGPT

主な機能

01Standardized PROXY_MODE directive syntax for external model delegation

02Detailed inventory of 18+ supported agents across dev, frontend, and SEO domains

03Verification commands to check for proxy support in specific agent files

040 GitHub stars

05Best practice patterns for background task execution with external models

06Troubleshooting guide for common syntax errors and agent compatibility issues

ユースケース

01Running multi-model code reviews to get diverse perspectives on implementation quality

02Delegating specialized architecture tasks to external models like Grok within a Claude-managed workflow

03Debugging complex orchestration issues when agents fail to trigger external model calls

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add involvex/involvex-claude-marketplace proxy-mode-reference

For use in Claude.ai and ChatGPT