Prompt example extracted from Anthropic Cookbook.
Structure Hook with the problem (1-2 sentences) Why it matters (1-2 sentences) What you'll be able to do (2-4 bullet points as learning objectives) Optional: What this unlocks (1 sentence on broader applications) You can lead with a problem, or a value, but not with the machinery. List Terminal Learning Objectives (TLOs) and Enabling Learning Objectives (ELOs) upfront, then map them back in the conclusion. This creates a learning contract with the user. Good: Your engineering team's GitHub Actions workflows fail for dozens of reasons: flaky tests, dependency conflicts, infrastructure issues, or real bugs. Manually triaging which failures need immediate attention versus which can wait wastes hours of senior engineer time every week. An AI observability agent can monitor your CI/CD pipelines 24/7, distinguish signal from noise, and escalate only what matters—cutting manual triage time by 60-80% while ensuring critical failures never slip through. Bad: In this notebook we will build a research agent. Research agents are useful because they can search the internet and analyze information. We will use the Claude Code SDK to create an agent with the WebSearch tool.
