This document will provide the best practices for setting up effective role-based prompts in AiSK. The goal is to ensure that AiSK can understand and perform tasks reliably, with a sharp focus on clarity, precision, and context.

1. Syntactical Optimization

• Maximize clarity: Focus on concise, telegraphic language. Every word should act as a trigger with deep meaning.
• Use high-impact keywords: Choose words related to the AiSK knowledge bases to maximize relevance.
• Avoid unnecessary language: Use dense instruction for input, ensuring the output remains smooth without sacrificing efficiency.

2. Role Setting Guidelines

a. State the Role Explicitly

• Direct role assignment: Clearly state the role at the start of the prompt to immediately narrow the model's focus.
  • Example:As a data scientist, your job is to explain how overfitting occurs in machine learning.

b. Align Role with Task

• Match the role precisely to the task to avoid ambiguity.
  • Example:As a project manager, your job is to create a thorough product plan for a 6-month product launch.

c. Specialize Roles for Nuanced Tasks

• Use specialized roles for complex tasks to provide precise instructions.
  • Example:As a cybersecurity expert, explain best practices for securing cloud environments.

d. Combine Roles if Needed

• Blend roles to cover multiple perspectives when necessary for tasks that span different domains.
  • Example:As both a marketer and UX designer, propose changes to boost conversion rates on a mobile app.

e. Contextualize the Role

• Add situational context to guide tone or approach.
  • Example:As a customer support agent, respond with empathy to a complaint about unexpected billing.

f. Avoid Over-Specification

• Keep the role defined yet flexible to avoid constraining AiSK bot too much.
  • Example:As a financial advisor, outline key strategies for retirement planning.

3. Instruction Structuring and Execution

a. Clear Instruction Delivery

• Short direct prompt: Use high-impact keywords to guide immediate task execution.
• Comprehensive documentation: Include a detailed task guide that AiSK can refer to nuanced details (e.g., persona, tone).

b. Granular Steps

• Break down complex steps into smaller, manageable actions to improve execution reliability.
• Use delimiters between instruction sets to avoid confusion or task overlap.

4. Ensure Consistency and Precision

a. Explicit Definitions

• Use few examples to define acceptable and unacceptable changes, improving consistency in the output.

b. Strengthen Instructions

• Reinforce important instructions via phrases like “check your work” or “take your time” to encourage thoroughness and minimize errors.

5. Specialized AiSK for Diverse Task Sets

• Create separate knowledge bases/AiSK bots for tasks requiring different areas of expertise. This will avoid overcomplicating a single AiSK bot and improve reliability.

6. Example-Based Instruction Reinforcement

a. Few-Shot Prompting for Clarity

• Use few-shot examples to clarify complex instructions or desired output, especially when retrieving critical data or performing complex actions.

b. Provide Good Output Examples

• Offer examples of ideal output to ensure AiSK understands expectations regarding the format, tone, and content.

Follow the above best practices to ensure that AiSK bot consistently delivers reliable, context-specific, and high-quality outputs tailored to each role and task. Optimizing both role setting and instruction structuring helps streamline complex operations while maintaining clarity and precision.