Skills

Skills are modular, reusable blocks of code that execute securely on ChatboxAI's infrastructure. They allow models to perform complex, multi-step actions without requiring your application to handle the intermediate code execution or network requests.

1. Skills vs. Tools vs. Connectors

To understand where Skills fit into the ChatboxAI ecosystem, it helps to compare them to our other capability extensions:

• Tools (Function Calling): The model asks your server to run code. Your server runs the code and sends the result back to the model.
• Connectors (MCP): ChatboxAI connects to a third-party server (like your internal Jira instance) to fetch data.
• Skills: ChatboxAI executes custom code directly on its own secure servers on behalf of the model.

If you have a workflow that is completely self-contained (e.g., crunching numbers, generating a chart, or pinging an open API), a Skill is the fastest and most efficient way to achieve it.

2. Using Built-in Skills

ChatboxAI provides several highly optimized built-in skills for common computing tasks. You attach them to a request via the skills array.

POST https://chatboxai.co.in/api/v1/responses

{
  "model": "chatboxai/llama-3.3-70b",
  "input": "Calculate the 100th Fibonacci number and generate a bar chart of the first 20.",
  "skills": ["python_interpreter"]
}

Popular Built-in Skills

• python_interpreter: Securely runs Python code. Ideal for math, data analysis, and generating visual plots.
• web_browser: A headless browser that can scrape dynamic single-page applications (SPAs) and navigate through websites.
• sql_executor: Connects to a read-only database replica to query large datasets.

3. Defining Custom Skills (Serverless Execution)

You can define custom JavaScript/TypeScript skills that execute on ChatboxAI's edge network. This allows you to build serverless workflows directly into your API calls.

Example: Custom Weather Skill

In this example, the model writes a JavaScript function that fetches weather data. ChatboxAI executes the code, and the model reads the result.

{
  "model": "chatboxai/claude-4.5-sonnet",
  "input": "What's the weather in Tokyo?",
  "skills": [
    {
      "name": "get_weather",
      "description": "Fetches current weather for a location.",
      "code": "async function run(args) { const res = await fetch(`https://api.weatherapi.com/v1/current.json?q=${args.location}&key=YOUR_KEY`); return await res.json(); }"
    }
  ]
}

Security & Sandboxing

All custom skill code is executed inside a heavily restricted WebAssembly (Wasm) sandbox to ensure safety and isolation.

• Maximum Execution Time: 10 seconds.
• Memory Limit: 128 MB.
• Allowed Modules: fetch, URL, standard JavaScript built-ins.
• File System: No local file system access.

4. Pricing & Limits

Using Skills requires additional compute resources.

• Built-in Skills: Tools like the python_interpreter carry a small per-invocation premium. Please check the Pricing page for details.
• Custom Skills: Custom JavaScript skills are currently included at no extra cost beyond standard token billing.
• Tier Availability: Custom skills and Built-in skills are available to both Free and Paid plans, though Free plan users are bound by the 300 requests/day global limit.