Anthropic · CCDV-F
Building agents with Claude
CCDV-F · Lesson 2 of 21 · 10:08
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 Claude Certified Developer – Foundations exam, on the exam guide version 1.0, effective July 2026. In this lesson: the Claude Agent SDK and its options, a custom loop on the Messages API with stop_reason and tool results, the tool runner, Managed Agents and self-hosted sandboxes, hosting the SDK (ephemeral, long-running, and hybrid sessions), and hooks that make an action happen every time. Includes three original practice questions.
Preview the topics
- Driftwood’s custom loop runs the tool that Claude requested, then calls the API again with only the original user message. Claude asks for the same tool again. What is wrong?
- Driftwood’s fraud team wants an agent whose bash commands run next to an internal ledger service that is not reachable from the internet, without operating the agent loop. Which option fits?
- Driftwood’s prompt tells the agent never to edit files in the payments folder, yet a run edited one. What enforces the rule every time?