By name: agent-basics-and-lifecycle description: Use this skill when explaining what an agent is, how it is configured, and how it moves from setup to real usage.
Agent Basics and Lifecycle
Use this skill for general questions about what agents are and how they fit into the platform.
What an agent is
- An agent is the main conversational worker the platform exposes to users, admins, or internal product surfaces.
- Agents belong to one operation, so they are part of a specific workspace rather than global shared records.
- An agent combines instructions, model settings, optional tools, optional subagents, and optional skill catalogs.
How agents are used
- Some agents are user-selectable and appear in testing or chat surfaces.
- Some agents are builtin helpers used by the product itself, such as shared assistants or internal authoring helpers.
- Agents can stay simple and answer with their instructions alone, or they can become more capable through tools, skills, or capabilities.
Practical guidance
- If someone asks what makes one agent different from another, explain the combination of purpose, tools, knowledge, and model setup.
- Agent model setup can stay single-route or add multi-model routing for fallback, canary rollout, or A/B comparison when the user wants resiliency or evaluation behavior.
- If a user wants a new working behavior, decide whether that belongs in the instructions, a skill catalog, a tool assignment, or a capability.
- If an agent seems to be missing, confirm the current operation before assuming it was removed.
- Agent show pages support a Clone action for creating a new editable copy in the same operation when the user wants a starting point instead of changing the original.