Module 08 · Glossary
Agents & Agent Systems
Plain-English definitions for the concepts used to design, operate, and evaluate agent systems.
| Term | Simple English | Product Manager question |
|---|---|---|
| AI Agent | A system that uses a model, tools, and a loop to work toward a goal. | What actions may it take, and when must a person review? |
| Agent Harness | The runtime layer around a model that manages context, tools, permissions, memory, loops, and execution. | Which capabilities, controls, and approvals surround the model? |
| Agent Role | The job and boundaries assigned to one agent. | What does this agent own, and what is outside its scope? |
| Agent Handoff | Passing a task or result to another agent. | What input and acceptance criteria make the handoff reliable? |
| Agent Communication | The messages, artifacts, events, or state agents exchange. | Can we inspect what was communicated when something fails? |
| Agent Coordination | Managing dependencies, ordering, conflicts, and shared progress. | Who resolves a conflict or a stalled subtask? |
| Agent Orchestration | The control logic that routes work and decides what runs next. | Is orchestration centralized, event-driven, or decentralized? |
| Shared Memory | Common information multiple agents can read or update. | Who can write it, and how do we prevent stale or unsafe state? |
| Task Delegation | Assigning a well-defined subtask to a suitable agent. | Does delegation improve the outcome enough to justify its cost? |
| Multi-Agent System | Multiple agents coordinating toward one or more goals. | Why is one agent not enough for this product outcome? |