Agent-based Automated Claim Matching with Instruction-following LLMs