AI and System Prompt Architecture: Layered Instruction Design

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-creators-foundations-AI-and-system-prompt-architecture-r11a4-creators

1. What problem does layered architecture solve when designing system prompts?

   1. It breaks down a large monolithic prompt into maintainable named sections
   2. It prevents AI models from generating harmful content
   3. It guarantees that the prompt will pass security reviews
   4. It automatically writes all test cases needed for evaluation

2. An AI assistant is asked to refactor a 1500-token system prompt into a layered architecture. How many layers should the resulting structure contain according to the recommended approach?

   1. As many layers as there are distinct tasks the prompt performs
   2. Four layers with explicit responsibilities
   3. Three layers: system, user, and assistant
   4. Two layers: instructions and examples

3. Which of the following is explicitly listed as something AI can do when helping architect system prompts?

   1. Replace iterative refinement with structure alone
   2. Draft per-layer responsibility comments
   3. Validate that the prompt meets all legal requirements
   4. Guarantee that modular prompts always perform better than monolithic ones

4. A developer refactors their system prompt using layered architecture but skips re-running their evaluation set. What risk does this create?

   1. The prompt may become too short to function properly
   2. The AI will automatically fix any errors introduced
   3. The layers will become misaligned with the token limit
   4. Silent quality drops may go unnoticed

5. In the context of layered prompt architecture, what is a 'foundation' layer likely to contain?

   1. Specific task instructions and output formats
   2. Core identity, values, and behavioral boundaries that apply to all interactions
   3. Code snippets demonstrating desired functionality
   4. Examples of ideal user-AI conversations

6. A team uses AI to refactor their system prompt into layers but finds that performance degrades after deployment. What should they have done differently?

   1. Used more layers in their architecture
   2. Avoided using AI for refactoring entirely
   3. Added more examples to each layer
   4. Re-run their evaluation set after each change

7. What does it mean for a system prompt to be 'monolithic' in this lesson's context?

   1. It is a single large block of text containing all instructions
   2. It enforces strict character limits on outputs
   3. It uses only a single AI model
   4. It has been validated by multiple independent reviewers

8. Which task is explicitly identified as something AI cannot do in prompt architecture?

   1. Draft per-layer responsibility comments
   2. Refactor a monolithic prompt into named layers
   3. Replace iterative refinement with structure alone
   4. Suggest test cases per layer

9. When AI suggests test cases per layer during refactoring, what is the purpose of these test cases?

   1. To document the history of changes made
   2. To increase the token count of the prompt
   3. To verify that each layer functions as intended after refactoring
   4. To provide examples for the AI to copy

10. Why might a developer choose layered architecture over a single large prompt even if it doesn't guarantee better performance?

    1. Layered prompts automatically pass all safety checks
    2. Layered prompts are easier to understand, debug, and modify incrementally
    3. Layered prompts always require fewer tokens
    4. Layered prompts work without any examples

11. What is the primary benefit of giving each layer an explicit responsibility in prompt architecture?

    1. It enables real-time modifications without evaluation
    2. It clarifies what each section should contain and makes the prompt easier to manage
    3. It allows the prompt to use more AI models simultaneously
    4. It automatically generates documentation for stakeholders

12. What does the lesson identify as a key risk of refactoring system prompts without evaluation?

    1. The AI may refuse to follow the restructured instructions
    2. Silent quality drops may occur without obvious signs
    3. The layers may exceed token limits
    4. The prompt may become too complex for the AI to process

13. In a 4-layer prompt architecture, where would task-specific output format instructions most appropriately be placed?

    1. In a task-specific layer dedicated to output handling
    2. Distributed equally across all four layers
    3. In a separate fifth layer not mentioned in the architecture
    4. In the foundation layer with core identity information

14. The lesson mentions that AI can 'refactor a monolithic prompt into named layers.' What does 'refactor' mean in this context?

    1. Add new capabilities to the prompt
    2. Compress the prompt to use fewer tokens
    3. Restructure the prompt's content without changing its external behavior
    4. Translate the prompt into another language

15. What distinguishes a well-architected layered prompt from a poorly structured one, according to the principles in this lesson?

    1. Clear, explicit responsibilities assigned to each named layer
    2. The use of AI to generate all layer content
    3. The inclusion of at least ten examples per layer
    4. The number of tokens in each layer