Merit AC
Domain 9 of 14 — Observability and evaluation Day 19 of 30

Connect a request to its full trace

The prompt

For one recent agent session, produce a single trace that connects the request to model activity, tool selection, policy decisions, tool results, latency, errors, and final outcome. What tooling would you need to make this a five-minute exercise instead of an hour of log archaeology?

Why it's built that way

Agent systems require traces that connect a request to model activity, tool selection, policy decisions, tool results, latency, errors, and final outcomes — AgentCore Observability's answer is CloudWatch plus OpenTelemetry-compatible tracing, but the requirement holds regardless of stack.

What to do with the answer

If this took more than a few minutes, that's your next observability investment, not a one-off exercise to repeat manually.

← Day 18All 30 daysDay 20 →