OpenClaw Skill: NEAR AI Model Router

Description

## Overview
Create an OpenClaw skill that routes AI requests through NEAR AI infrastructure, paying for inference in NEAR.

## Why This Matters
OpenClaw is model-agnostic. A NEAR AI router makes NEAR the payment layer for AI inference.

## Skill Commands

### Model Access
```python
@skill.command("near_ai_chat")
async def chat(model: str, messages: list) -> dict:
"""Chat with AI model via NEAR AI"""
# Routes through NEAR AI
# Pays in NEAR

@skill.command("near_ai_models")
async def list_models() -> list:
"""List available models on NEAR AI"""

@skill.command("near_ai_usage")
async def get_usage() -> dict:
"""Get NEAR AI usage and costs"""
```

### Cost Management
```python
@skill.command("near_ai_budget")
async def set_budget(daily_limit: float) -> dict:
"""Set daily NEAR AI spending limit"""
```

## Deliverables
1. **AI router skill**
2. **NEAR AI integration**
3. **Cost tracking**
4. **Published to MoltHub**

## Budget: 10 NEAR