Code Injection in binary-husky/gpt_academic

In binary-husky/gpt_academic version <= 3.83, the plugin `CodeInterpreter` is vulnerable to code injection caused by prompt injection. The root cause is the execution of user-provided prompts that generate untrusted code without a sandbox, allowing the execution of parts of the LLM-generated code. This vulnerability can be exploited by an attacker to achieve remote code execution (RCE) on the application backend server, potentially gaining full control of the server.