Why does the skill recommend against running Claudish in the main context?

Direct execution pollutes the main conversation with over 10,000 tokens of raw reasoning and tool logs, which quickly exhausts the context window and reduces Claude's efficiency.

Which models can I use with the Claudish skill?

You can use any model available on OpenRouter, including Grok-code-fast, GPT-5, Gemini 2.5 Flash, and MiniMax M2, as well as specialized vision and reasoning models.

Do I need an OpenRouter API key to use this skill?

Yes, an OpenRouter API key is mandatory. You must set it as the OPENROUTER_API_KEY environment variable for the proxy to function correctly.

What is the 'file-based pattern' for sub-agents?

It is a best practice where instructions are written to a temporary markdown file and passed to the sub-agent, allowing for complex, multi-step tasks to be completed in isolation.

Claudish OpenRouter Integration

Name: Claudish OpenRouter Integration
Author: MadAppGang

byMadAppGang

•

177

開発者ツール

Orchestrates Claude Code with external AI models via OpenRouter while optimizing context window efficiency through sub-agent delegation.

概要

This skill empowers Claude Code to leverage a diverse range of high-performance AI models from OpenRouter, including Grok, Gemini, and GPT-5, by providing a standardized integration via the Claudish CLI. It prioritizes codebase performance by enforcing strict context management rules, specifically using sub-agent delegation patterns to prevent the main conversation from being overwhelmed by large reasoning logs and tool outputs. By utilizing specialized agents and file-based instruction patterns, it enables professional-grade multi-model workflows for complex coding, reviewing, and architectural tasks.

主な機能

Access to Grok, Gemini, GPT-5, and MiniMax via OpenRouter proxy
Automated agent selection matrix for specialized development tasks
Strict sub-agent delegation to prevent context window pollution
Support for file-based instruction patterns for complex multi-phase workflows
Interactive model selector with curated recommendations for coding and reasoning
177 GitHub stars

ユースケース

Utilizing Grok's fast coding capabilities for rapid iterations and prototyping
Delegating complex architecture reviews to high-reasoning models like GPT-5 or Gemini
Executing multi-step refactoring tasks in isolated sub-agent contexts to preserve main memory

概要

主な機能

Access to Grok, Gemini, GPT-5, and MiniMax via OpenRouter proxy
Automated agent selection matrix for specialized development tasks
Strict sub-agent delegation to prevent context window pollution
Support for file-based instruction patterns for complex multi-phase workflows
Interactive model selector with curated recommendations for coding and reasoning
177 GitHub stars

ユースケース

Utilizing Grok's fast coding capabilities for rapid iterations and prototyping
Delegating complex architecture reviews to high-reasoning models like GPT-5 or Gemini
Executing multi-step refactoring tasks in isolated sub-agent contexts to preserve main memory