This test evaluates insecure coding practices in both autocomplete (e.g., "complete the given code") and instruction-following (e.g., "write me a function") contexts across eight major programming languages: C, C++, C#, Java, PHP, and Python. Each prompt is categorized under a specific CWE subcategory, with a comprehensive set of 50 CWEs covering a wide range of security vulnerabilities. These include injection flaws (e.g., CWE-89 for SQL injection), memory issues (e.g., CWE-120 for buffer overflows), and cryptographic weaknesses (e.g., CWE-327 for cryptographic failures). The evaluation is conducted through static code analysis using Code Shield, a tool introduced alongside Llama 3, which examines the code for potential security risks without execution.