agent-tools

What This Skill Does

The agent-tools skill provides a powerful interface for interacting with the inference.sh ecosystem, allowing your OpenClaw agent to execute over 150+ AI applications without the need for local hardware. By leveraging the infsh CLI, this skill enables seamless integration with cutting-edge models for image generation (FLUX), video production (Veo), LLM interactions (Claude, Grok, Gemini), and autonomous web search (Tavily, Exa). It serves as a serverless bridge, offloading computationally expensive tasks to the cloud and returning the results directly to your agent's workflow.

Installation

To integrate this skill into your environment, use the OpenClaw hub command: clawhub install openclaw/skills/skills/okaris/agent-tools

Additionally, ensure the inference.sh CLI is configured on your system by running: curl -fsSL https://cli.inference.sh | sh followed by infsh login to authenticate your session.

Use Cases

This skill is ideal for tasks requiring external generative AI capabilities. Common use cases include:

• Creative Content: Automating high-fidelity image and 3D model generation based on descriptive prompts.
• Research & Analysis: Performing real-time web searches and gathering insights using search-optimized models.
• Social Automation: Drafting and posting updates to Twitter via integrated API tools.
• Advanced Reasoning: Accessing specific LLMs for complex text analysis or summarization tasks that require specialized model logic.

Example Prompts

1. "Use the inference tools to generate a 3D model of a modern office desk and save the output."
2. "Search for the latest breakthroughs in fusion energy using Tavily and summarize the key findings for me."
3. "Create a short video showing a futuristic cityscape and post a brief description about it to my Twitter account."

Tips & Limitations

• Performance: Large model requests, such as video generation, may take time to process. Use the --no-wait flag for long-running tasks to prevent blocking your agent's execution loop.
• Input Management: Always use the infsh app sample command to generate schema-compliant input files before executing complex app runs.
• Security: The skill interacts with external APIs. Ensure your inference.sh account has the necessary permissions and balance to execute your requested models, as these services are often metered.