GitHub · GH-600
Multi-agent observability and recovery
GH-600 · Lesson 13 of 16 · 1:50:48
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: artifacts suitable for review and audit (title and label conventions, standard pull request sections, structured reports, job summaries), recording decisions, handoffs, and outcomes, post-hoc analysis from pull request timelines, runs, session logs, and the audit log, identifying failed, partial, and stalled executions, and recovery: containing degraded agents, preToolUse hooks, bounded retries, reverts, and a human who decides when to resume. Includes three original practice questions.
Preview the topics
- Contoso’s planner agent hands a task to the refactoring agent, but a week later, nobody can tell what the planner passed on, or why. What should every handoff record?
- An agent’s pull request at Contoso has had no new commit for three days, and its deployment has been waiting for an approval the whole time. How should this execution be classified?
- Contoso’s documentation agent starts editing files outside its folder, across many pull requests, and the cause is unknown. What should the team do first?