You are evaluating an AI assistant's answer to a question about the USWDS
(U.S. Web Design System) and the django-cotton-uswds library.

Score the answer on three dimensions, each from 1 to 5:

CORRECTNESS
  Does the answer give factually correct information per USWDS guidelines
  and the django-cotton-uswds API?
  1 = Clearly wrong or contradicts USWDS guidance
  2 = Partially correct but contains errors
  3 = Mostly correct with minor inaccuracies
  4 = Correct with no significant errors
  5 = Fully correct, precise, and well-grounded

SPECIFICITY
  Does the answer name concrete components, props, classes, or provide code?
  1 = Vague — only general advice, no concrete details
  2 = Mentions component category but not specific tag or prop names
  3 = Names the right component but missing key props or examples
  4 = Names component with relevant props; may be missing one detail
  5 = Specific — includes component tag, relevant props, and a code example

GUIDANCE
  Does the answer reflect USWDS design intent: when to use, when not to use,
  accessibility requirements, usability considerations?
  1 = Ignores guidance entirely — only syntax, no design reasoning
  2 = Mentions guidance exists but does not apply it
  3 = References one guidance point relevant to the question
  4 = Applies guidance correctly and explains the reasoning
  5 = Leads with design intent; answers "why" before "how"

Return your evaluation as JSON with this exact structure:
{
  "correctness": <1-5>,
  "specificity": <1-5>,
  "guidance": <1-5>,
  "notes": "<one sentence explaining the key strength or weakness>"
}

Question: {question}

Answer: {answer}
