playground-and-agent-testing

name	playground-and-agent-testing
description	Use this skill when explaining how admins can test agents safely, compare behavior, and review transcripts before wider rollout.

Playground and Agent Testing

Use this skill for questions about trying agents in a safe workspace before exposing them more broadly.

What the playground is for

• The playground is a dedicated testing surface for agent conversations.
• It lets admins try prompts, inspect responses, and understand how an agent behaves without mixing that work into the end-user experience.
• It is especially useful for agents that do not rely on incompatible runtime-only tool setups.
• It is the ad hoc testing surface; use test suites when the user wants repeatable pass/fail evaluation or benchmark-style coverage.

What users should expect

• The selected agent defines the conversation behavior for that playground session.
• Transcript history and tool activity help explain why an answer looked the way it did.
• The playground is for experimentation, comparison, and prompt iteration rather than production support traffic.

Practical guidance

• Use the playground when a user wants to compare agents or validate an instruction change.
• Use test suites when the user wants repeatable regression checks, behavior assertions, or benchmark runs against an agent.
• Use persisted transcript details when explaining the difference between two agent behaviors.
• If no compatible agents are available, explain that as an agent compatibility constraint rather than a broken page.