llm-models

What This Skill Does

The llm-models skill integrates the power of over 100 state-of-the-art Large Language Models into your OpenClaw environment. By leveraging the inference.sh CLI and the OpenRouter ecosystem, this skill provides a unified gateway to premium intelligence from Anthropic (Claude 4.5 family), Google (Gemini 3), Kimi, GLM, and more. It abstracts the complexity of managing multiple API keys and provider-specific endpoints, allowing you to access advanced reasoning, code generation, and content creation capabilities via a single, consistent interface. Whether you need high-speed responses for conversational tasks or the deep reasoning capabilities of frontier models, this skill handles the heavy lifting, including automatic fallback mechanisms and cost-optimized routing.

Installation

To begin, ensure you have the OpenClaw environment configured. Install the skill using the following command: clawhub install openclaw/skills/skills/okaris/llm-models

Once installed, the skill relies on the inference.sh CLI. Initialize it by running: curl -fsSL https://cli.inference.sh | sh && infsh login

This setup connects your local terminal to the global model network securely. The install process is transparent, lightweight, and requires no elevated permissions.

Use Cases

• Software Engineering: Generate boilerplate code, refactor legacy scripts, write unit tests, or debug complex logic using high-reasoning models like Claude Opus 4.5.
• Content Creation: Draft emails, blog posts, or marketing copy with specific tones or structures, leveraging the balanced performance of Claude Sonnet 4.5.
• Agentic Workflows: Integrate models into automated pipelines where the skill performs reasoning or data summarization based on system-provided context.
• Quick Research: Extract insights from large blocks of text or perform rapid translations across multiple languages.

Example Prompts

1. "Use Claude Sonnet to explain the architectural differences between microservices and monoliths, focusing on scalability."
2. "Generate a Python function that uses regular expressions to extract email addresses from a raw text file and run it through the Kimi K2 agent for logic verification."
3. "Translate the following project requirement document into French, ensuring professional tone and technical accuracy."

Tips & Limitations

• Model Selection: Always use any-model if you are unsure of your requirements, as it optimizes for cost and performance. Use Claude Opus specifically for high-complexity tasks that require deep logical chaining.
• System Prompts: Always utilize the system field in your input JSON when you need to define a consistent persona (e.g., 'You are a senior DevOps engineer').
• Data Privacy: Remember that inputs are sent to the OpenRouter API. Do not send highly sensitive credentials or proprietary keys through the prompt without local obfuscation.
• Cost Awareness: While the interface simplifies usage, be mindful of token usage with larger models like Claude Opus, which are more expensive than Haiku.