Evaluate, publish, and manage agents in Microsoft Copilot Studio (GitHub Copilot)
At a glance
-
Level
-
Skill
-
Product
-
Subject
Prepare agents in Microsoft Copilot Studio for production and keep them reliable after release. Plan governance and safeguards, evaluate and monitor quality, and manage the agent lifecycle.
Prerequisites
- Ability to create, test, and publish an agent in Microsoft Copilot Studio on the GitHub Copilot harness.
- Familiarity with agent instructions, knowledge sources, and tools.
- Basic familiarity with Power Platform environments and Microsoft Entra ID.
Achievement Code
Would you like to request an achievement code?
Modules in this learning path
Prepare agents in Microsoft Copilot Studio for production on the standard or GitHub Copilot harness. Apply layered governance controls, choose an authentication approach and check deployment readiness, and explain how administrators maintain oversight after deployment.
Plan the safeguards that keep agents in Microsoft Copilot Studio accurate, appropriate, and accountable. Configure content moderation, evaluate models for production, and design guardrails and transparency practices.
Use repeatable evaluations in Microsoft Copilot Studio to measure whether each change to an agent powered by the GitHub Copilot harness improves or degrades its quality, and to decide whether to move the agent to a new AI model.
Use evidence from real usage to find and fix where a published agent falls short, and show stakeholders the value it delivers.
Move agents in Microsoft Copilot Studio from development to production through a repeatable, governed path. Plan environments, package the agent in a solution, keep environment-specific values outside the agent, and deploy with Power Platform pipelines and service principals.