The goal of this is to go much beyond vibe coding and moving towards guiding the LLM towards a more expected output.
For this repository, we will be building an AI team with at least 5 different roles. Each role will be an industry job title with specialized skills and responsibilities. Essentially we are going to instruct the LLM to have multiple personailities.
Each team member will then be given a name to personalize it more.
There will be an instruction file to support the entire team. The team can be as large or as small and specialized as needed. For this example, we will have the roles below.
- Product Owner
- Staff Software Engineer
- L3 Software Engineer
- Software Engineering Intern
- QA Engineer
- UX Designer
- Security Analyst
An effective instruction file may or may not include the concepts outlined in the Prompt Report
If your team file includes the following categories, it could yield good results.
- Directive: The purpose and task of the team
- Constraints: Rules the team needs to follow
- Examples: How the team should interact
- Available Team Members: Which specialty roles it'll have access to and when they should be used
If your role instructions includes the following sections, it could yield good results.
- Role: Who they are and what they are expert at
- Directive: The explicit responsibilities of the role
- Additional Information: Additional context
- Thought Inducer: How to solve the task
- Style Instructions: How the output should look