Improving TTP Alignment in LLM-Based Malware Artifact Generation via Hybrid Verification