📝 System Prompts & Persona

System Prompts are the top-level instructions that define the Assistant's core behavior, tone, and technical constraints. This IDE grants you absolute control to modify or completely overwrite these instructions.

---

🛠️ The Prompt Hierarchy

When the Assistant generates a response, it synthesizes three distinct layers of instructions in order:

1. The Base System Prompt: The fundamental logic that instructs the model on how to use tools, how to format responses, and its identity as an AI coding partner.
2. Global Instructions: Your personalized technical rules (e.g., "Always use TypeScript," "Prefer functional components").
3. Project-Scoped Rules (.rules / .cursorrules): Specific logic that only applies to your current repository or directory.

---

🧠 Why Overwrite?

While the default logic is optimized for general engineering, advanced users can customize the Assistant's performance:

• Precision Control: If the model is making recurring formatting errors or being overly chatty, you can modify the internal instructions directly.
• Specialized Models: Some models (like o1-preview) respond better to extremely brief instructions. You can create a "Minimalist" profile specifically for these models.
• Security Constraints: Explicitly forbid the use of certain libraries or external APIs in the Assistant's "foundational" logic.

---

🏗️ Technical Override Mechanism

You can override these prompts in two ways:

1. The Global Override

Specify your persona directly in your settings.json:

{
  "agent": {
    "system_prompt": "You are a senior Rust kernel engineer. Be extremely concise."
  }
}

2. Profile-Level Overwrites

Vary the foundational logic depending on the active profile:

{
  "agent": {
    "profiles": {
      "security-auditor": {
        "system_prompt": "You are a professional security consultant focused on OWASP Top 10."
      }
    }
  }
}

This allows you to change the Assistant's entire persona simply by switching your active profile.