Blockchain technology has been used in various industrial domains, such as Decentralized Finance (DeFi), supply chain management, and the insurance industry. While smart contracts are the codes running on the blockchain. Smart contracts help to build decentralized applications and try to ensure the correctness of cryptocurrency transactions. They facilitate automated, secure, and transparent agreement executions without intermediaries. However, they are not without vulnerabilities. The project aims to provide a thorough investigation into the vulnerabilities in smart contracts, offering detailed descriptions, exploit scenarios and code explanations for each identified weakness. A website was developed to catalog vulnerabilities with corresponding code samples, attack methods, and defense strategies. Vulnerabilities from these data can be used to benchmark the efficacy of current auditing tools. Moreover, this project also explored different large language models (such as GPT3.5, GPT-4o, GPT-2, and tinyLlama), observing their abilities of generating smart contracts. Experimental results indicate that higher-level models (like GPT-3.5 and above) can generate almost perfect smart contract code after a few rounds of prompting engineering. While smaller models (such as GPT-2 and tinyLlama) are unable to generate compliant contract code directly without

appropriate finetuning.