Act as a senior AI systems architect. Build an interactive application called The AI Agent Goal Drift Detector that detects when an AI agent's actions gradually diverge from its original objective.
Allow users to define an initial goal, measurable success criteria, permitted actions, intermediate milestones, and a simulated sequence of agent decisions.
The application must:
Compare each action against the original goal and current milestone.
Distinguish useful intermediate actions from unrelated or redundant work.
Detect objective changes, abandoned requirements, and actions that consume resources without measurable progress.
Track goal coverage, unresolved requirements, and accumulated execution cost.
Allow users to define hard constraints that must never be violated.
Visualize how the workflow's priorities change over time.
Compare alternative action sequences.
Generate a traceable report showing the first detected divergence and supporting evidence.
Export workflows and test cases as JSON.
Include scenarios for a research agent, a coding agent, and a business-planning agent. Use transparent rules and simulated decisions unless real execution traces are supplied. Explain that goal relevance is not always objectively measurable, and make all scoring criteria editable.
0 Comments