[gitlab-org/gitlab] Fix Pipeline flow: move instructions to system prompt

### ROOT CAUSE The current implementation uses user prompts for system-level instructions, which can lead to ambiguity and unintended behavior from the AI. Instructions should be separated from user input to ensure clarity and consistency in the AI's responses. ### CODE FIX 1. **Separate System and User Prompts**: Move all system-level instructions to the system prompt, reserving the user prompt for dynamic input. 2. **Update Prompt Structure**: Restructure the prompt to explicitly distinguish between system instructions and user input. **Example Fix** (assuming a ChatGPT-like API structure): ```python def generate_response(system_instructions, user_input): messages = [ {"role": "system", "content": system_instructions}, {"role": "user", "content": user_input} ] response = model.generate(messages=messages) return response ``` **Changes**: - Replace any hard-coded instructions in user prompts with the system prompt. - Ensure all AI behavior guidelines are passed as system-level content. This change improves clarity and ensures the AI adheres to predefined instructions consistently.