How are external models invoked in Claude Code?

External models are invoked via the claudish CLI using Bash commands, which provides a deterministic and 100% reliable execution path compared to standard LLM completions.

Which external LLM providers are supported?

It supports a wide variety of models including xAI (Grok), Google (Gemini), OpenAI (GPT), MiniMax, and DeepSeek through the claudish interface.

Can the skill remember which models I prefer for specific tasks?

Yes, it automatically saves and reuses model preferences per context (e.g., research, coding, or debugging) in a local configuration file.

How do I ensure an external model execution was successful?

The skill includes a verification protocol to check exit codes, output byte size, and stderr logs to confirm valid responses were received.

Is parallel execution supported for multi-model tasks?

Yes, by using the run_in_background parameter in orchestration messages, you can trigger multiple external models to run simultaneously.

External Model Task Orchestration

Name: External Model Task Orchestration
Author: MadAppGang

byMadAppGang

•

230

•

データサイエンスとML

Integrates external LLMs like Grok, Gemini, and GPT into Claude Code workflows using reliable Bash-based orchestration.

This skill provides a standardized framework for invoking non-Claude models within orchestration workflows. By leveraging the claudish CLI, it ensures deterministic and reliable model execution, automated context-based preference saving, and parallel processing capabilities. It is ideal for developers who need to perform cross-model validation, offload specialized tasks to domain-specific LLMs, or conduct automated multi-model code reviews without leaving the Claude environment.

主な機能

01Parallel execution patterns for high-performance orchestration

02Standardized verification protocols for exit codes and output quality

03230 GitHub stars

04Extensive support for providers including xAI, Google, OpenAI, and DeepSeek

05Deterministic external model invocation via Bash and claudish CLI

06Context-aware preference storage for automatic model reuse based on task type

ユースケース

01Automated benchmarking of model responses for research and debugging queries

02Offloading specific sub-tasks to specialized models during complex multi-agent builds

03Running cross-model code reviews to compare logic and optimization suggestions

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add madappgang/claude-code task-external-models

For use in Claude.ai and ChatGPT

Download Skill