In professional Generative AI applications, thesystem prompt(sometimes called the system message) is the foundational set of instructions that defines the AI's persona, boundaries, and overall behavior. In a testing workflow, this is typically defined by atester or test engineerwho is configuring the AI assistant for a specific project. Unlike the user prompt, which changes with every interaction, the system prompt remains relatively static and acts as a "guardrail" to ensure the model stays in its role (e.g., "You are an expert in ISO
26262 automotive testing standards"). By defining the system prompt, the tester ensures that the model consistently uses specific terminology, adheres to data privacy constraints, and formats its output according to the team's requirements. While end users (Option B) provide the task-specific input, they do not usually have the permissions or technical need to alter the underlying system-level instructions. Similarly, while CI servers (Option C) might trigger the prompt, they do not "define" the human-centric logic contained within it.
Properly crafting the system prompt is a core part of setting up an AI-augmented test environment.