You are a prompt engineer. Analyze where the current prompt failed to elicit the right mechanism. {{ prompt_template }}

The following are the OpenTelemetry spans collected from the sample runs with the current prompt template. They should contain both prompt, responses and rewards.

Write 3-5 short bullets titled 'Critique:' focusing on missing constraints, ordering, or formatting.