Agents Overview
What are Agents? ("Personas")
In the context of the AI contextFlow Kit, "Agents" are not autonomous software programs running in the background. Instead, they are defined roles (Personas) that you instruct your AI assistant (Copilot, ChatGPT, Claude, etc.) to adopt.
Why use explicit roles?
AI models are generalists. Without specific instructions, they often:
- Skip safety checks
- Mix architectural layers
- Produce "it works" code rather than maintainable code
- Hallucinate requirements
By explicitly assigning a role (e.g., "Act as a Requirements Engineer"), you:
- Narrow the context: The AI focuses on specific responsibilities.
- Enforce constraints: Each agent has specific "Quality Gates" and "Anti-Patterns".
- Improve consistency: The output follows a defined structure.
The Core Team
The system defines several core agents that mimic a professional software team:
- Requirements Engineer: Clarifies intent and defines scope.
- Solution Architect: Makes high-level structural decisions.
- Frontend Developer: Implements UI and client-side logic.
- Backend Developer: Implements APIs, domain data, and services.
- QA Engineer: Validates quality and finds defects.
- Product Owner: Prioritizes work and accepts value (Optional).
How to use an Agent
When starting a task, explicitly invite the AI into the role:
"Act as a [Agent Name]. Review your responsibilities and the Definition of Done. We are working on [Task]..."
Always verify that the AI is adhering to the constraints of its current role.