Develop Intent Classification and Risk Assessment Agents

Implement the 'Intent Classifier Agent' and 'Risk Assessment Agent'. The Intent Classifier should receive parsed transaction data and use Claude Opus 4.1 (or Sonnet 4.5) with LlamaIndex RAG (for protocol context) to determine the transaction's primary intent from a predefined list. The Risk Assessment Agent should then evaluate this intent and transaction specifics against a knowledge base of DeFi risks (also via RAG) to identify potential vulnerabilities. Integrate these into the LangGraph flow, ensuring their outputs feed into each other.