$ paw-test check suite.yaml
exit code: 1
--- stdout ---
Running paw.test check on: golden_date_normalizer
(<WORKDIR.......................................................>/quoted.paw) (backend: MockPAWBackend)
  [FAIL] Input: 'today' -> '"2026-09-11"'
         regex_match: Output '"2026-09-11"' does not match pattern '^(\d{4}-\d{2}-\d{2}|INVALID)$'
         expected: got "2026-09-11" want 2026-09-11
  [FAIL] Input: 'new year' -> '"2026-01-01"'
         regex_match: Output '"2026-01-01"' does not match pattern '^(\d{4}-\d{2}-\d{2}|INVALID)$'
         expected: got "2026-01-01" want 2026-01-01
  [FAIL] Input: 'February 30th' -> '"INVALID"'
         regex_match: Output '"INVALID"' does not match pattern '^(\d{4}-\d{2}-\d{2}|INVALID)$'
Correct against expected: 0/2 (0.0%) -- all 2 scored
  2 of 2 standard cases carry an answer key.
  Of the ones not counted correct, 2 would match if a JSON string quote were stripped -- the adapter wraps its answers
in quotes. Reporting only: a quoted output is a real defect for any consumer of the adapter, so the scored number above
stays 0/2.

Pass rate: 0.0% (0/3) -- all 3 scored (backend: MockPAWBackend)
