Autonomy, artifacts, and human intervention
GH-600 · Lesson 3 of 16 · 17:28
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the GH-600 exam (GitHub Certified: Agentic AI Developer). In this lesson: planning the degree of autonomy from risk and growing it behind enforced guardrails, the built-in protections of Copilot cloud agent (one branch, draft pull requests, no approval or merge, Approve and run workflows, firewall, CodeQL and secret scanning), inspectable artifacts (pull request sections, signed commits linked to session logs, workflow artifacts and job summaries), and human intervention at decision points: steering and stopping sessions, fast reviews, and automatic workflow runs only where the risk is low. Includes three original practice questions.
Preview the topics
- A Contoso developer assigns an issue to Copilot cloud agent, and then wants to approve the agent’s pull request alone, to save time. What happens?
- During an audit, Contoso must explain why the cloud agent changed a refund rule three weeks ago. Where should the auditor start?
- Halfway through a session, a Contoso reviewer sees that the cloud agent is writing its own error handler instead of using the shared one. What should the reviewer do?