{
  "test_case_id": "A11Y-QC-002",
  "title": "Version-frozen calibration set",
  "category": "quality_control",
  "procedure_version": "2026.08.27-rc.9",
  "rule_type": "quality_control",
  "wcag_criteria": [],
  "level": "advisory",
  "risk": "medium",
  "risk_basis": "Detect inconsistent procedure interpretation and unreliable decisions before production findings support claims. People most likely affected: people affected by assessment decisions whose reliability depends on consistent, evidence-based procedure execution. Uncalibrated reviewers may reach conflicting results, hide uncertainty or create misleading accessibility claims. In the recorded accessibility-quality-control environment, use frozen fixtures, independent scoring and documented reconciliation; unresolved assessor disagreement can turn the same evidence into inconsistent client findings.",
  "review_status": "draft",
  "reviewed_at": "",
  "review_note": "Template content is subject to a hash-bound internal consistency review. Formal assessment claims require competent execution, current evidence and the applicable legal, participant and organisational authority.",
  "method_label": "Quality-control review",
  "tester_level": "Accessibility quality lead",
  "estimated_duration": "Typically 125–181 minutes for the stated single-target assumptions; use the component breakdown when estimating a real sample.",
  "execution_class": "manual",
  "automation_support": "none",
  "simulation_policy": "simulation_may_support_but_not_replace_manual_execution",
  "plain_summary": "Have assigned reviewers independently score version-frozen examples, measure agreement and resolve differences without rewriting their original decisions.",
  "objective": "Detect inconsistent procedure interpretation and unreliable decisions before production findings support claims.",
  "test_subject": "An identified execution-agent cohort applying one frozen procedure version to a version-frozen calibration suite.",
  "test_targets": [
    "Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture."
  ],
  "applicability": "Apply before execution agents use a new or materially revised procedure for production assessment and at the governance-defined recalibration interval.",
  "applicability_questions": [
    "Is the procedure version frozen?",
    "Does the calibration suite cover all five outcomes plus material exceptions and borderline cases?",
    "Were expected decisions separately reviewed and kept hidden from assigned reviewers during scoring?",
    "Are agreement thresholds and review rules fixed before results are known?"
  ],
  "inapplicability_conditions": [
    "No execution-agent cohort is assigned to use the procedure. This does not make the procedure itself reliable or approved."
  ],
  "assumptions": [
    "The selected scope identifies the product version, technologies, states and important user journeys being evaluated.",
    "The tester knows how to use this method and any required assistive technology.",
    "Known browser or assistive-technology limitations are recorded instead of being assigned to the product without evidence."
  ],
  "accessibility_support_assumptions": [
    "The result covers only the browsers, platforms, readers and assistive technologies that were actually tested.",
    "Do not apply a result from one technology combination to an untested combination or to every user."
  ],
  "prerequisites": [
    "The product area and test targets are agreed before testing starts.",
    "The operating system, browser, viewport and input method are recorded.",
    "Test data is safe to use and does not expose unnecessary personal information."
  ],
  "required_environment_fields": [
    "catalogue_version",
    "review_protocol",
    "review_protocol_version",
    "calibration_set",
    "calibration_set_version",
    "evidence_set_reference",
    "reviewer_count"
  ],
  "steps": [
    {
      "step_id": "S1",
      "title": "Freeze the calibration package",
      "phase": "prepare",
      "action": "Record the catalogue, procedure and fixture versions; expected outcome key; exception and borderline coverage; assigned execution-agent IDs and versions; distinct verification-agent ID and protocol; conflicts; agreement metric; threshold; and rerun rule before revealing any answers.",
      "test_target": "Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture.",
      "required_state": "Immutable calibration package before execution-agent assignment.",
      "tool_or_command": "Version control, checksums and a blinded answer-key store.",
      "expected": "The package and decision thresholds cannot be changed after results are known without a new versioned run.",
      "evidence_to_record": "Versions/checksums, coverage matrix, agent IDs and versions, conflicts, metric formula, threshold and answer-key integrity record.",
      "decision_effect": "An unfrozen package makes the calibration untested.",
      "recovery_or_repeat": "Create a successor version rather than altering the active package.",
      "novice_guidance": "The expected answer key must pass a separate hash-bound agent verification; unresolved normative interpretation remains cantTell.",
      "scenario_purpose": "For Version-frozen calibration set, prepare the named scenario: Freeze the calibration package. Target: Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture. The preparation must establish this result: The package and decision thresholds cannot be changed after results are known without a new versioned run.",
      "user_impact": "Uncalibrated reviewers may reach conflicting results, hide uncertainty or create misleading accessibility claims. This scenario isolates that risk in the action named Freeze the calibration package.",
      "failure_observation": "During Version-frozen calibration set, the target, starting state, environment or scope record remains ambiguous for the scenario named Freeze the calibration package. The preparation result cannot be demonstrated: The package and decision thresholds cannot be changed after results are known without a new versioned run.",
      "decision_rules": {
        "passed": "Record passed for the Version-frozen calibration set scenario named Freeze the calibration package only when the evidence (Versions/checksums, coverage matrix, agent IDs and versions, conflicts, metric formula, threshold and answer-key integrity record.) supports: The package and decision thresholds cannot be changed after results are known without a new versioned run.",
        "failed": "Record failed when the action is completed and this failure observation is reproduced: During Version-frozen calibration set, the target, starting state, environment or scope record remains ambiguous for the scenario named Freeze the calibration package. The preparation result cannot be demonstrated: The package and decision thresholds cannot be changed after results are known without a new versioned run.",
        "not_applicable": "No execution-agent cohort is assigned to use the procedure.",
        "blocked": "Record blocked, not passed, when The target is applicable, but a required state, environment, action or evidence item was not completed."
      },
      "limitations_and_claim_boundary": "This scenario records the action named Freeze the calibration package for Version-frozen calibration set in accessibility-quality-control. It does not establish behaviour for another target, state or environment and does not set a WCAG conformance decision."
    },
    {
      "step_id": "S2",
      "title": "Collect blinded independent decisions",
      "phase": "test",
      "action": "Have every execution agent execute every assigned fixture independently using the frozen procedure. Record target outcome, evidence, rationale, exception use and time only where the plan requires it. Do not expose expected or peer answers.",
      "test_target": "Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture.",
      "required_state": "Identical fixture and procedure versions for every execution agent.",
      "tool_or_command": "Normal execution worksheet with blinded answer access.",
      "expected": "Original decisions are complete, independently produced and immutable.",
      "evidence_to_record": "Agent/fixture IDs, versions, outcomes, evidence, rationale, exception decisions and completion timestamp.",
      "decision_effect": "Missing fixtures are untested; shared state or revealed answers invalidates the run.",
      "recovery_or_repeat": "Preserve originals and schedule a new blinded rerun after corrective action.",
      "novice_guidance": "Never replace the first decision with the corrected answer; calibration measures the original independent judgement.",
      "scenario_purpose": "For Version-frozen calibration set, execute the named scenario: Collect blinded independent decisions. Target: Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture. Compare the observation with this result: Original decisions are complete, independently produced and immutable.",
      "user_impact": "Uncalibrated reviewers may reach conflicting results, hide uncertainty or create misleading accessibility claims. This scenario isolates that risk in the action named Collect blinded independent decisions.",
      "failure_observation": "After the tester completes the Version-frozen calibration set scenario named Collect blinded independent decisions, the recorded observation for the target is inconsistent with the required result: Original decisions are complete, independently produced and immutable.",
      "decision_rules": {
        "passed": "Record passed for the Version-frozen calibration set scenario named Collect blinded independent decisions only when the evidence (Agent/fixture IDs, versions, outcomes, evidence, rationale, exception decisions and completion timestamp.) supports: Original decisions are complete, independently produced and immutable.",
        "failed": "Record failed when the action is completed and this failure observation is reproduced: After the tester completes the Version-frozen calibration set scenario named Collect blinded independent decisions, the recorded observation for the target is inconsistent with the required result: Original decisions are complete, independently produced and immutable.",
        "not_applicable": "No execution-agent cohort is assigned to use the procedure.",
        "blocked": "Record blocked, not passed, when The target is applicable, but a required state, environment, action or evidence item was not completed."
      },
      "limitations_and_claim_boundary": "This scenario records the action named Collect blinded independent decisions for Version-frozen calibration set in accessibility-quality-control. It does not establish behaviour for another target, state or environment and does not set a WCAG conformance decision."
    },
    {
      "step_id": "S3",
      "title": "Calculate agreement and error patterns",
      "phase": "analyse",
      "action": "Compare independent decisions with the separately reviewed answer key and with each other using the predeclared metric. Report confusion among all five outcomes, exception disagreements and evidence-quality failures.",
      "test_target": "Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture.",
      "required_state": "All original records locked.",
      "tool_or_command": "Reproducible calculation or analysis script with saved inputs and output.",
      "expected": "The metric, threshold comparison and material disagreement patterns are reproducible.",
      "evidence_to_record": "Input record hashes, calculation version, metric result, confusion matrix, disagreements and threshold decision.",
      "decision_effect": "A below-threshold or irreproducible result fails the control.",
      "recovery_or_repeat": "Correct calculation defects transparently with a versioned report; do not alter source decisions.",
      "novice_guidance": "A single headline percentage can hide systematic confusion between failed, cantTell and untested outcomes.",
      "scenario_purpose": "For Version-frozen calibration set, execute the named scenario: Calculate agreement and error patterns. Target: Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture. Compare the observation with this result: The metric, threshold comparison and material disagreement patterns are reproducible.",
      "user_impact": "Uncalibrated reviewers may reach conflicting results, hide uncertainty or create misleading accessibility claims. This scenario isolates that risk in the action named Calculate agreement and error patterns.",
      "failure_observation": "After the tester completes the Version-frozen calibration set scenario named Calculate agreement and error patterns, the recorded observation for the target is inconsistent with the required result: The metric, threshold comparison and material disagreement patterns are reproducible.",
      "decision_rules": {
        "passed": "Record passed for the Version-frozen calibration set scenario named Calculate agreement and error patterns only when the evidence (Input record hashes, calculation version, metric result, confusion matrix, disagreements and threshold decision.) supports: The metric, threshold comparison and material disagreement patterns are reproducible.",
        "failed": "Record failed when the action is completed and this failure observation is reproduced: After the tester completes the Version-frozen calibration set scenario named Calculate agreement and error patterns, the recorded observation for the target is inconsistent with the required result: The metric, threshold comparison and material disagreement patterns are reproducible.",
        "not_applicable": "No execution-agent cohort is assigned to use the procedure.",
        "blocked": "Record blocked, not passed, when The target is applicable, but a required state, environment, action or evidence item was not completed."
      },
      "limitations_and_claim_boundary": "This scenario records the action named Calculate agreement and error patterns for Version-frozen calibration set in accessibility-quality-control. It does not establish behaviour for another target, state or environment and does not set a WCAG conformance decision."
    },
    {
      "step_id": "S4",
      "title": "Verify causes and corrective actions",
      "phase": "review",
      "action": "Have the distinct verification agent inspect each material disagreement against the frozen method and evidence. Classify method ambiguity, source dispute, capability gap, fixture defect or evidence failure and assign corrective action without inventing consensus; use cantTell when the evidence does not support a cause.",
      "test_target": "Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture.",
      "required_state": "Original decisions and calculation locked and available to the verification agent.",
      "tool_or_command": "Hash-bound agent-verification record.",
      "expected": "Every material disagreement has an evidence-supported resolution or remains an open accuracy dispute.",
      "evidence_to_record": "Disagreement, evidence, classification, decision, verifier ID/protocol, owner, due date and affected procedure/fixture version.",
      "decision_effect": "An unresolved material accuracy dispute prevents a passed quality-control result.",
      "recovery_or_repeat": "Version procedure or fixtures when corrective action changes their meaning, then rerun.",
      "novice_guidance": "Post-scoring agreement is not the original inter-rater agreement and must remain separate.",
      "scenario_purpose": "For Version-frozen calibration set, execute the named scenario: Verify causes and corrective actions. Target: Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture. Compare the observation with this result: Every material disagreement has an evidence-supported resolution or remains an open accuracy dispute.",
      "user_impact": "Uncalibrated reviewers may reach conflicting results, hide uncertainty or create misleading accessibility claims. This scenario isolates that risk in the action named Verify causes and corrective actions.",
      "failure_observation": "After the tester completes the Version-frozen calibration set scenario named Verify causes and corrective actions, the recorded observation for the target is inconsistent with the required result: Every material disagreement has an evidence-supported resolution or remains an open accuracy dispute.",
      "decision_rules": {
        "passed": "Record passed for the Version-frozen calibration set scenario named Verify causes and corrective actions only when the evidence (Disagreement, evidence, classification, decision, verifier ID/protocol, owner, due date and affected procedure/fixture version.) supports: Every material disagreement has an evidence-supported resolution or remains an open accuracy dispute.",
        "failed": "Record failed when the action is completed and this failure observation is reproduced: After the tester completes the Version-frozen calibration set scenario named Verify causes and corrective actions, the recorded observation for the target is inconsistent with the required result: Every material disagreement has an evidence-supported resolution or remains an open accuracy dispute.",
        "not_applicable": "No execution-agent cohort is assigned to use the procedure.",
        "blocked": "Record blocked, not passed, when The target is applicable, but a required state, environment, action or evidence item was not completed."
      },
      "limitations_and_claim_boundary": "This scenario records the action named Verify causes and corrective actions for Version-frozen calibration set in accessibility-quality-control. It does not establish behaviour for another target, state or environment and does not set a WCAG conformance decision."
    },
    {
      "step_id": "S5",
      "title": "Verify corrective action with a rerun",
      "phase": "test",
      "action": "After versioned capability, method or fixture changes, run a new blinded calibration set that includes equivalent but unrevealed examples. Recalculate the same predeclared metric.",
      "test_target": "Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture.",
      "required_state": "Corrective action completed; new frozen package and hidden answers.",
      "tool_or_command": "Same controlled calibration workflow.",
      "expected": "The cohort meets the threshold without an unresolved material dispute.",
      "evidence_to_record": "New package/version, capability or method change, independent decisions, metric and comparison with the prior run.",
      "decision_effect": "A failed rerun remains failed; do not waive it through schema validation or organisational preference.",
      "recovery_or_repeat": "Repeat only under a new versioned run with a documented reason.",
      "novice_guidance": "Reusing revealed examples tests memorisation, not reliable application of the method.",
      "scenario_purpose": "For Version-frozen calibration set, execute the named scenario: Verify corrective action with a rerun. Target: Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture. Compare the observation with this result: The cohort meets the threshold without an unresolved material dispute.",
      "user_impact": "Uncalibrated reviewers may reach conflicting results, hide uncertainty or create misleading accessibility claims. This scenario isolates that risk in the action named Verify corrective action with a rerun.",
      "failure_observation": "After the tester completes the Version-frozen calibration set scenario named Verify corrective action with a rerun, the recorded observation for the target is inconsistent with the required result: The cohort meets the threshold without an unresolved material dispute.",
      "decision_rules": {
        "passed": "Record passed for the Version-frozen calibration set scenario named Verify corrective action with a rerun only when the evidence (New package/version, capability or method change, independent decisions, metric and comparison with the prior run.) supports: The cohort meets the threshold without an unresolved material dispute.",
        "failed": "Record failed when the action is completed and this failure observation is reproduced: After the tester completes the Version-frozen calibration set scenario named Verify corrective action with a rerun, the recorded observation for the target is inconsistent with the required result: The cohort meets the threshold without an unresolved material dispute.",
        "not_applicable": "No execution-agent cohort is assigned to use the procedure.",
        "blocked": "Record blocked, not passed, when The target is applicable, but a required state, environment, action or evidence item was not completed."
      },
      "limitations_and_claim_boundary": "This scenario records the action named Verify corrective action with a rerun for Version-frozen calibration set in accessibility-quality-control. It does not establish behaviour for another target, state or environment and does not set a WCAG conformance decision."
    },
    {
      "step_id": "S6",
      "title": "Repeat material variations and recovery",
      "phase": "test",
      "test_target": "Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture.",
      "required_state": "For Version-frozen calibration set: Passed, failed, inapplicable, cantTell and untested outcomes. Material exceptions, borderline cases and at least one evidence-quality failure where meaningful.",
      "tool_or_command": "For Version-frozen calibration set, use the same recorded method, then change only the state, environment or recovery condition being compared.",
      "action": "Repeat the Version-frozen calibration set action for each inventoried variation that can change the outcome. Include relevant default, revealed, error, interruption, recovery and confirmation states and complete the selected process from start to finish where this test can affect task completion.",
      "expected": "Every material variation and required recovery path has its own reproducible observation; no unexecuted state is merged into a passing result.",
      "failure_observation": "After the tester completes the Version-frozen calibration set scenario named Repeat material variations and recovery, the recorded observation for the target is inconsistent with the required result: Every material variation and required recovery path has its own reproducible observation; no unexecuted state is merged into a passing result.",
      "evidence_to_record": "Variation identifier; changed state or environment; start and end points; actual observation; recovery result; intervention; evidence reference and target outcome.",
      "decision_effect": "For Version-frozen calibration set, a reproduced mismatch fails the affected target; an omitted required variation remains untested; conflicting evidence remains cannot tell.",
      "recovery_or_repeat": "Repeat this action for every required variation of Run the version-frozen execution-team calibration and after a material target, content, state, environment or configuration change. Preserve the original observation.",
      "novice_guidance": "For Version-frozen calibration set, change one material condition at a time where practical. Do not assume that a passing default state covers errors, overlays, orientation changes, interruptions or recovery.",
      "limitations_and_claim_boundary": "This scenario records the action named Repeat material variations and recovery for Version-frozen calibration set in accessibility-quality-control. It does not establish behaviour for another target, state or environment and does not set a WCAG conformance decision.",
      "scenario_purpose": "For Version-frozen calibration set, execute the named scenario: Repeat material variations and recovery. Target: Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture. Compare the observation with this result: Every material variation and required recovery path has its own reproducible observation; no unexecuted state is merged into a passing result.",
      "user_impact": "Uncalibrated reviewers may reach conflicting results, hide uncertainty or create misleading accessibility claims. This scenario isolates that risk in the action named Repeat material variations and recovery.",
      "decision_rules": {
        "passed": "Record passed for the Version-frozen calibration set scenario named Repeat material variations and recovery only when the evidence (Variation identifier; changed state or environment; start and end points; actual observation; recovery result; intervention; evidence reference and target outcome.) supports: Every material variation and required recovery path has its own reproducible observation; no unexecuted state is merged into a passing result.",
        "failed": "Record failed when the action is completed and this failure observation is reproduced: After the tester completes the Version-frozen calibration set scenario named Repeat material variations and recovery, the recorded observation for the target is inconsistent with the required result: Every material variation and required recovery path has its own reproducible observation; no unexecuted state is merged into a passing result.",
        "not_applicable": "No execution-agent cohort is assigned to use the procedure.",
        "blocked": "Record blocked, not passed, when The target is applicable, but a required state, environment, action or evidence item was not completed."
      }
    },
    {
      "step_id": "S7",
      "title": "Record the scoped quality decision",
      "phase": "record",
      "action": "A distinct competent reviewer records pass, fail, cantTell or untested for this cohort and procedure version, the production-use limitation and next calibration date. Preserve all source records and keep any organisational release decision separate.",
      "test_target": "Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture.",
      "required_state": "All required review findings and reruns complete.",
      "tool_or_command": "Quality-control register and immutable evidence references.",
      "expected": "The decision is scoped to identified reviewers, procedure version, suite version and time period.",
      "evidence_to_record": "Reviewer identity, method/version, decision time, scope, limitations, source-record references, next review and status effect.",
      "decision_effect": "A passed control supports team readiness only; it does not approve a product claim or organisational release.",
      "recovery_or_repeat": "Recalibrate after material source, procedure, platform, adapter or reviewer change.",
      "novice_guidance": "The reviewer cannot create legal clearance, participant evidence, conformance or organisational approval.",
      "scenario_purpose": "For Version-frozen calibration set, complete the record for the scenario named Record the scoped quality decision and preserve evidence that demonstrates this result: The decision is scoped to identified reviewers, procedure version, suite version and time period.",
      "user_impact": "Uncalibrated reviewers may reach conflicting results, hide uncertainty or create misleading accessibility claims. This scenario isolates that risk in the action named Record the scoped quality decision.",
      "failure_observation": "The Version-frozen calibration set evidence or decision record for the scenario named Record the scoped quality decision does not demonstrate the required result: The decision is scoped to identified reviewers, procedure version, suite version and time period.",
      "decision_rules": {
        "passed": "Record passed for the Version-frozen calibration set scenario named Record the scoped quality decision only when the evidence (Reviewer identity, method/version, decision time, scope, limitations, source-record references, next review and status effect.) supports: The decision is scoped to identified reviewers, procedure version, suite version and time period.",
        "failed": "Record failed when the action is completed and this failure observation is reproduced: The Version-frozen calibration set evidence or decision record for the scenario named Record the scoped quality decision does not demonstrate the required result: The decision is scoped to identified reviewers, procedure version, suite version and time period.",
        "not_applicable": "No execution-agent cohort is assigned to use the procedure.",
        "blocked": "Record blocked, not passed, when The target is applicable, but a required state, environment, action or evidence item was not completed."
      },
      "limitations_and_claim_boundary": "This scenario records the action named Record the scoped quality decision for Version-frozen calibration set in accessibility-quality-control. It does not establish behaviour for another target, state or environment and does not set a WCAG conformance decision."
    }
  ],
  "expectations": [
    {
      "expectation_id": "E1",
      "statement": "Procedure, fixture, answer key, cohort, metric and threshold are frozen before scoring.",
      "outcome_if_unmet": "untested"
    },
    {
      "expectation_id": "E2",
      "statement": "Original independent decisions and evidence are preserved and produce a reproducible agreement result.",
      "outcome_if_unmet": "cantTell"
    },
    {
      "expectation_id": "E3",
      "statement": "The threshold is met with no unresolved material accuracy dispute and a distinct agent-verification decision is recorded.",
      "outcome_if_unmet": "failed"
    }
  ],
  "exceptions": [
    "Use only an exception stated in the relevant standard, or an agreed out-of-scope decision. Record the reason and evidence.",
    "A tool limitation, lack of time or unavailable assistive technology is a limitation or incomplete result, not a pass."
  ],
  "outcome_definitions": {
    "passed": "The applicable test target meets every expectation in this rule or procedure.",
    "failed": "The applicable test target does not meet at least one expectation in this rule or procedure.",
    "inapplicable": "The test subject contains no target that matches this rule or procedure's applicability.",
    "cantTell": "The tester cannot fully determine applicability or whether every expectation is met.",
    "untested": "The test subject or target was not evaluated using this rule or procedure."
  },
  "examples": [
    {
      "example_id": "A11Y-QC-002-EX-PASSED",
      "outcome": "passed",
      "scenario": "Representative calibration example, not client evidence. Procedure: Version-frozen calibration set. Target: Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture. Action: Record the catalogue, procedure and fixture versions; expected outcome key; exception and borderline coverage; assigned execution-agent IDs and versions; distinct verification-agent ID and protocol; conflicts; agreement metric; threshold; and rerun rule before revealing any answers. Evidence: Versions/checksums, coverage matrix, agent IDs and versions, conflicts, metric formula, threshold and answer-key integrity record. Observed result: The package and decision thresholds cannot be changed after results are known without a new versioned run."
    },
    {
      "example_id": "A11Y-QC-002-EX-FAILED",
      "outcome": "failed",
      "scenario": "Representative calibration example, not client evidence. Procedure: Version-frozen calibration set. Target: Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture. The tester completes the recorded action, but the captured evidence does not demonstrate: The package and decision thresholds cannot be changed after results are known without a new versioned run. The failed target and state remain separate from unexecuted targets."
    },
    {
      "example_id": "A11Y-QC-002-EX-INAPPLICABLE",
      "outcome": "inapplicable",
      "scenario": "Representative scope example, not client evidence. No execution-agent cohort is assigned to use the procedure. The completed inventory is retained as the basis for the inapplicable outcome."
    },
    {
      "example_id": "A11Y-QC-002-EX-CANTTELL",
      "outcome": "cantTell",
      "scenario": "Representative unresolved example, not client evidence. The answer key, metric inputs, independence or cause of material disagreement cannot be validated. The record identifies the conflicting evidence and the expertise or comparison needed to resolve it."
    },
    {
      "example_id": "A11Y-QC-002-EX-UNTESTED",
      "outcome": "untested",
      "scenario": "Representative execution-gap example, not client evidence. The target is applicable, but a required state, environment, action or evidence item was not completed. The missing action, state or environment remains open and is not converted to a pass."
    }
  ],
  "sampling_rule": "Apply this control to the preselected assessment sample. Do not change the sample or threshold after results are known.",
  "complete_process_requirement": "If this check is part of a user journey, test the selected journey from start to finish, including relevant errors and recovery. Record any path that was not tested.",
  "pass_criteria": "Every assigned execution agent completes the frozen suite independently, the predeclared agreement threshold is met, exception and evidence decisions are defensible, no material accuracy dispute remains and the distinct verification agent records the scoped decision.",
  "fail_criteria": "The cohort misses the predeclared threshold, a material systematic error remains, independence is compromised or corrective action fails verification.",
  "evidence_requirements": [
    "Catalogue/procedure/fixture versions and hashes, reviewed answer-key custody, cohort, conflicts, metric and threshold.",
    "Every auditor's original outcomes, evidence, rationale and timestamps.",
    "Inputs, calculation version, metric, confusion matrix and threshold comparison.",
    "Disagreement classification, evidence, decision, owner, changes and rerun result.",
    "Distinct reviewer identity, method/version, decision time, scope, limitations and next calibration date."
  ],
  "official_source_ids": [
    "W3C-WCAG-EM-1",
    "W3C-WCAG-EM-2",
    "W3C-ACT-RULES-FORMAT"
  ],
  "act_rule_mapping": {
    "format_alignment": "This is a project-authored manual procedure, not a W3C Approved ACT Rule.",
    "approved_rule_ids": [],
    "approved_rules": [],
    "mapping_status": "No supporting ACT rule selected",
    "registry_reviewed_at": "2026-08-27",
    "registry_source_id": "W3C-ACT-RULES",
    "mapping_rationale": "An ACT result can support only the part of this procedure covered by that rule. It does not replace the remaining manual or assistive-technology checks."
  },
  "technique_traceability": [],
  "platform_ids": [
    "accessibility-quality-control"
  ],
  "review_requirement": "A qualified accessibility reviewer must confirm the scope, evidence and result before it is reported as final.",
  "known_issues": [
    "Check relevant browser, tool or assistive-technology release notes before assigning unexpected behaviour to the product.",
    "If responsibility is unclear, record the reproducible behaviour and use cannot tell until it is resolved."
  ],
  "change_log": [
    {
      "version": "2026.08.27-rc.9",
      "change": "Reworked for plain language, case-focused steps, scoped evidence and transparent draft review status."
    },
    {
      "version": "2026.08.27-rc.9",
      "change": "Case-specific specialist content, evidence and platform boundaries completed for hash-bound agent verification."
    },
    {
      "version": "2026.08.27-rc.9",
      "change": "Individually authored reference-procedure content prepared for competent review and execution."
    },
    {
      "version": "2026.08.27-rc.9",
      "change": "Removed administrative filler steps, added contextual risk and result boundaries, and prepared the concise worksheet flow for internal consistency review."
    }
  ],
  "tags": [
    "calibration",
    "distinct-quality-review",
    "quality-control",
    "quality_control"
  ],
  "status": "draft",
  "authoring_readiness": "content_complete",
  "short_title": "Run the version-frozen execution-team calibration",
  "catalogue_version": "2026.08.27-rc.9",
  "identity": {
    "stable_case_id": "A11Y-QC-002",
    "procedure_title": "Version-frozen calibration set",
    "short_user_centred_title": "Run the version-frozen execution-team calibration",
    "procedure_version": "2026.08.27-rc.9",
    "catalogue_version": "2026.08.27-rc.9",
    "status": "draft",
    "category": "quality_control",
    "procedure_type": "quality_control",
    "execution_class": "manual",
    "intended_tester_expertise": "Accessibility quality lead",
    "effort_model": {
      "model": "component_effort_v2",
      "small_scope_minutes": {
        "minimum": 125,
        "maximum": 181
      },
      "warning": "Add a new per-target execution and evidence allowance for each additional target class or state that could produce a different result. Add setup time for each distinct environment.",
      "calculation_guidance": "Add the recorded setup, per-target execution, evidence and technical-review components; then apply the stated variation rule.",
      "unit": "minutes",
      "setup": {
        "typical": 30,
        "includes": "Confirm scope and target, prepare safe data, establish the required state and record the tool or environment version."
      },
      "per_target_execution": {
        "typical": 45,
        "includes": "Perform the procedure once for one target in its selected states; this excludes remediation and retesting."
      },
      "evidence_capture": {
        "typical": 20,
        "includes": "Capture, name and cross-reference the minimum reproducible evidence while removing unnecessary personal or secret data."
      },
      "technical_review": {
        "typical": 30,
        "includes": "Check applicability, observations, outcome logic, evidence sufficiency and report wording before issue publication."
      },
      "typical_total_range": {
        "minimum": 125,
        "maximum": 181
      },
      "assumptions": "one frozen review set and one reviewer pair; disagreements, adjudication and reruns are additional effort",
      "variation": "Add a new per-target execution and evidence allowance for each additional target class or state that could produce a different result. Add setup time for each distinct environment.",
      "excludes": "Client coordination, inaccessible-environment recovery, recruitment, legal review, remediation implementation and waiting time unless the procedure states otherwise."
    },
    "related_procedures": [],
    "replaces": [],
    "replaced_by": []
  },
  "purpose_and_user_impact": {
    "barrier_detected": "Detect inconsistent procedure interpretation and unreliable decisions before production findings support claims.",
    "people_likely_affected": [
      "People affected by assessment decisions whose reliability depends on consistent, evidence-based procedure execution."
    ],
    "functionality_or_information_at_risk": "Uncalibrated reviewers may reach conflicting results, hide uncertainty or create misleading accessibility claims.",
    "result_scope": "quality_control_evidence",
    "claim_boundary": "This procedure determines whether Run the version-frozen execution-team calibration meets its stated expectations for the recorded targets and accessibility-quality-control. Unexecuted targets, states and environments remain outside that result.",
    "why_it_matters": "Uncalibrated reviewers may reach conflicting results, hide uncertainty or create misleading accessibility claims. The procedure is necessary because detect inconsistent procedure interpretation and unreliable decisions before production findings support claims and the result can otherwise conceal a material barrier for the affected people named here."
  },
  "requirement_mappings": [
    {
      "source_id": "W3C-WCAG-EM-1",
      "requirement_id": "W3C-WCAG-EM-1",
      "requirement_level": "informative",
      "requirement_type": "advisory",
      "coverage_type": "supporting",
      "canonical_source": "https://www.w3.org/TR/WCAG-EM/",
      "source_status": "group_note",
      "source_version_or_publication_date": "WCAG-EM 1.0; W3C Note 10 July 2014",
      "date_verified": "2026-08-27",
      "reviewer_notes": "Supporting source. Procedure-level content review remains recorded separately.",
      "limitations": "This source does not override WCAG or create an additional WCAG conformance requirement.",
      "dependent_fields": [
        "prerequisites",
        "environment",
        "steps"
      ]
    },
    {
      "source_id": "W3C-WCAG-EM-2",
      "requirement_id": "W3C-WCAG-EM-2",
      "requirement_level": "informative",
      "requirement_type": "advisory",
      "coverage_type": "supporting",
      "canonical_source": "https://www.w3.org/TR/wcag-em-2/",
      "source_status": "draft_note",
      "source_version_or_publication_date": "WCAG-EM 2.0; Draft Note 5 February 2026",
      "date_verified": "2026-08-27",
      "reviewer_notes": "Supporting source. Procedure-level content review remains recorded separately.",
      "limitations": "This source does not override WCAG or create an additional WCAG conformance requirement.",
      "dependent_fields": [
        "prerequisites",
        "environment",
        "steps"
      ]
    },
    {
      "source_id": "W3C-ACT-RULES-FORMAT",
      "requirement_id": "W3C-ACT-RULES-FORMAT",
      "requirement_level": "informative",
      "requirement_type": "advisory",
      "coverage_type": "supporting",
      "canonical_source": "https://www.w3.org/TR/act-rules-format/",
      "source_status": "recommendation",
      "source_version_or_publication_date": "ACT Rules Format 1.1; W3C Recommendation 2026",
      "date_verified": "2026-08-27",
      "reviewer_notes": "Supporting source. Procedure-level content review remains recorded separately.",
      "limitations": "This source does not override WCAG or create an additional WCAG conformance requirement.",
      "dependent_fields": [
        "prerequisites",
        "environment",
        "steps"
      ]
    }
  ],
  "applicability_model": {
    "test_subject": "An identified execution-agent cohort applying one frozen procedure version to a version-frozen calibration suite.",
    "test_targets": [
      "Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture."
    ],
    "objective_questions": [
      "Is the procedure version frozen?",
      "Does the calibration suite cover all five outcomes plus material exceptions and borderline cases?",
      "Were expected decisions separately reviewed and kept hidden from assigned reviewers during scoring?",
      "Are agreement thresholds and review rules fixed before results are known?"
    ],
    "subjective_judgements": [
      "A distinct verification agent classifies whether disagreements expose ambiguous method wording, capability gaps, source disputes or fixture defects. Unsupported classification remains cantTell."
    ],
    "inapplicability_conditions": [
      "No execution-agent cohort is assigned to use the procedure. This does not make the procedure itself reliable or approved."
    ],
    "scope_exclusions": [
      "Procedures, versions, fixture versions and execution agents not included in the frozen calibration run."
    ],
    "required_states_and_variations": [
      "Passed, failed, inapplicable, cantTell and untested outcomes.",
      "Material exceptions, borderline cases and at least one evidence-quality failure where meaningful."
    ],
    "complete_process_dependencies": "Complete independent scoring, metric calculation, a distinct internal quality review, corrective action and a controlled rerun before a passed quality-control result.",
    "conforming_alternate_version_considerations": "Not applicable as a calibration shortcut; each procedure version requires its own governed calibration decision."
  },
  "normative_exceptions": [],
  "normative_exceptions_absent_reason": "This project quality-control method has no normative exception. Any permitted exclusion must be defined in the frozen quality plan and reported as a coverage limitation.",
  "environment": {
    "required_tester_expertise": "Accessibility quality lead",
    "required_platforms": [
      "accessibility-quality-control"
    ],
    "minimum_environment_fields": [
      "catalogue_version",
      "review_protocol",
      "review_protocol_version",
      "calibration_set",
      "calibration_set_version",
      "evidence_set_reference",
      "reviewer_count"
    ],
    "product_and_content_version_required": true,
    "viewport_and_orientation_required": true,
    "input_mode_required": true,
    "relevant_settings": [
      "Catalogue and procedure version",
      "Frozen fixtures/sample",
      "Assigned roles",
      "Predeclared threshold",
      "Calculation version"
    ],
    "known_platform_limitations": [
      "Quality-control evidence measures method reliability; it is not product accessibility or conformance evidence.",
      "Check relevant browser, tool or assistive-technology release notes before assigning unexpected behaviour to the product.",
      "If responsibility is unclear, record the reproducible behaviour and use cannot tell until it is resolved."
    ],
    "clean_state_requirements": [
      "Complete a known-good calibration in the same environment before product testing.",
      "Start from a recorded clean or known state and preserve any prior preferences, extensions, cache or profile settings that can affect output."
    ],
    "test_data_and_privacy_constraints": [
      "Use authorised synthetic or controlled data.",
      "Do not record credentials, secrets or unnecessary personal data."
    ],
    "coverage_boundary": "Quality-control evidence measures method reliability; it is not product accessibility or conformance evidence."
  },
  "decision_logic": {
    "target_passed": "Every assigned execution agent completes the frozen suite independently, the predeclared agreement threshold is met, exception and evidence decisions are defensible, no material accuracy dispute remains and the distinct verification agent records the scoped decision.",
    "target_failed": "The cohort misses the predeclared threshold, a material systematic error remains, independence is compromised or corrective action fails verification.",
    "target_inapplicable": "No execution-agent cohort is assigned to use the procedure.",
    "target_cant_tell": "The answer key, metric inputs, independence or cause of material disagreement cannot be validated.",
    "target_untested": "The target is applicable, but a required state, environment, action or evidence item was not completed.",
    "aggregation": "Preserve outcomes per execution agent and fixture. Apply only the predeclared metric and threshold; cantTell and untested prevent a passed control.",
    "wcag_requirement_decision": "This control makes no WCAG decision. It supports method reliability and execution-agent readiness within its recorded scope.",
    "conforming_alternate_version_effect": "Apply a conforming alternate version only after its WCAG definition and every conformance condition have been verified and recorded; otherwise it does not change the decision.",
    "passed_automation_follow_up": "Deterministic calculation and agent verification may decide the recorded quality-control scope but cannot create legal clearance, participant evidence, WCAG conformance or organisational approval.",
    "platform_responsibility": "When responsibility is unclear, preserve the reproducible behaviour and environment evidence and use cantTell until a qualified reviewer resolves it.",
    "claim_readiness": "A complete result may become ready_for_review. Schema validity, a passed target or a tool result cannot make it ready_for_scoped_claim without the independent evidence and governance gates."
  },
  "evidence_specification": [
    {
      "evidence_id": "EV1",
      "method": "frozen_package",
      "record": "Catalogue/procedure/fixture versions and hashes, reviewed answer-key custody, cohort, conflicts, metric and threshold.",
      "review_purpose": "Prove precommitment and version control.",
      "privacy_and_security": "Use a redacted extract or evidence reference. Do not include credentials, secrets or unnecessary personal data.",
      "required_for_formal_result": true
    },
    {
      "evidence_id": "EV2",
      "method": "independent_source_records",
      "record": "Every auditor's original outcomes, evidence, rationale and timestamps.",
      "review_purpose": "Review independence and preserve pre-adjudication decisions.",
      "privacy_and_security": "Use a redacted extract or evidence reference. Do not include credentials, secrets or unnecessary personal data.",
      "required_for_formal_result": true
    },
    {
      "evidence_id": "EV3",
      "method": "agreement_calculation",
      "record": "Inputs, calculation version, metric, confusion matrix and threshold comparison.",
      "review_purpose": "Reproduce reliability measures.",
      "privacy_and_security": "Use a redacted extract or evidence reference. Do not include credentials, secrets or unnecessary personal data.",
      "required_for_formal_result": true
    },
    {
      "evidence_id": "EV4",
      "method": "adjudication_and_action",
      "record": "Disagreement classification, evidence, decision, owner, changes and rerun result.",
      "review_purpose": "Review causes and verified remediation.",
      "privacy_and_security": "Use a redacted extract or evidence reference. Do not include credentials, secrets or unnecessary personal data.",
      "required_for_formal_result": true
    },
    {
      "evidence_id": "EV5",
      "method": "quality_review",
      "record": "Distinct reviewer identity, method/version, decision time, scope, limitations and next calibration date.",
      "review_purpose": "Document the scoped internal quality decision without representing independent verification.",
      "privacy_and_security": "Use a redacted extract or evidence reference. Do not include credentials, secrets or unnecessary personal data.",
      "required_for_formal_result": true
    }
  ],
  "calibration_examples": [
    {
      "example_id": "A11Y-QC-002-EX-PASSED",
      "outcome": "passed",
      "scenario": "The blinded execution-agent cohort meets the predeclared threshold on a frozen suite, no material dispute remains and a distinct verification agent records the scoped readiness decision.",
      "explanation": "All quality-control gates are met for the identified cohort and version.",
      "fixture_id": "",
      "fixture_version": "",
      "machine_testable": false
    },
    {
      "example_id": "A11Y-QC-002-EX-FAILED",
      "outcome": "failed",
      "scenario": "Auditors systematically mark blocked executions as cantTell target outcomes and the verified rerun still misses the threshold.",
      "explanation": "The control fails because outcome architecture remains unreliable.",
      "fixture_id": "",
      "fixture_version": "",
      "machine_testable": false
    },
    {
      "example_id": "A11Y-QC-002-EX-INAPPLICABLE",
      "outcome": "inapplicable",
      "scenario": "No auditor is assigned to use this procedure in production.",
      "explanation": "The control has no cohort target; the procedure is not thereby approved.",
      "fixture_id": "",
      "fixture_version": "",
      "machine_testable": false
    },
    {
      "example_id": "A11Y-QC-002-EX-CANTTELL",
      "outcome": "cantTell",
      "scenario": "The answer key's independent review record is missing, so agreement with it cannot establish correctness.",
      "explanation": "The evidence cannot support a defensible quality decision.",
      "fixture_id": "",
      "fixture_version": "",
      "machine_testable": false
    },
    {
      "example_id": "A11Y-QC-002-EX-UNTESTED",
      "outcome": "untested",
      "scenario": "The suite is frozen, but one assigned auditor has not completed the cases.",
      "explanation": "The cohort calibration remains incomplete.",
      "fixture_id": "",
      "fixture_version": "",
      "machine_testable": false
    }
  ],
  "review_record": {
    "author": "",
    "method_reviewer": "",
    "independent_standards_reviewer": "",
    "accessibility_ui_reviewer": "",
    "author_date": "",
    "method_review_date": "",
    "standards_review_date": "",
    "ui_review_date": "",
    "source_versions_reviewed": [],
    "exceptions_reviewed": false,
    "known_limitations": [
      "This is a source-checked reference implementation subject to independent method and standards agent verification; formal approval remains separate."
    ],
    "open_issues": [
      "No production calibration run, live execution evidence, agreement metric or organisational release decision is claimed.",
      "The calibration record is an internal quality control, not independent verification; production release authority remains separate."
    ],
    "decision": "pending",
    "next_review_date": "",
    "status_transition_reason": "",
    "release_gate_id": "",
    "release_gate_completed_at": "",
    "approval_authority": "",
    "approval_date": ""
  },
  "privacy_and_security": {
    "browser_storage": "Worksheet data is stored only in this browser's local storage until the user exports or clears it.",
    "network_transmission": "The static library does not upload worksheet data.",
    "data_minimisation": "Record only evidence necessary to reproduce and review the decision.",
    "redaction": "Remove credentials, secrets and unnecessary personal or participant data before export or sharing.",
    "retention_and_deletion": "Apply the engagement retention schedule and clear local browser data when it is no longer required."
  },
  "version_history": [
    {
      "version": "2026.08.27-rc.9",
      "change": "Reworked for plain language, case-focused steps, scoped evidence and transparent draft review status."
    },
    {
      "version": "2026.08.27-rc.9",
      "change": "Migrated to professional catalogue schema v3 with separated outcomes, structured authority, applicability, evidence, decision and review records."
    },
    {
      "version": "2026.08.27-rc.9",
      "change": "Specialist procedure mapped to its task family and current first-party platform guidance and checked for internal consistency; competent review and real execution remain separate."
    },
    {
      "version": "2026.08.27-rc.9",
      "change": "Replaced migrated boilerplate with a case-specific reference implementation and completed independent agent content verification."
    },
    {
      "version": "2026.08.27-rc.9",
      "change": "Added explicit preparation and variation/recovery steps plus a bounded representative worked example for paid-template use."
    },
    {
      "version": "2026.08.27-rc.9",
      "change": "Procedure guidance and presentation simplified without changing draft governance status."
    }
  ],
  "migration": {
    "from_schema": "accessibility-test-catalog.v2",
    "from_procedure_version": "2026.08.04.1",
    "migrated_at": "2026-08-27",
    "content_review_state": "commercial_completeness_upgrade_agent_verified",
    "stable_identifier_preserved": true,
    "legacy_url_preserved": true,
    "delivery_model": "professional_delivery_v2",
    "agent_content_validation": {
      "status": "complete",
      "method": "case_specific_semantic_and_operational_checks_v1",
      "checked_fields": [
        "target",
        "state",
        "tool",
        "action",
        "expected_observation",
        "failure_observation",
        "evidence",
        "outcome_rules",
        "claim_boundary",
        "compact_examples"
      ],
      "human_approval_effect": "none"
    }
  },
  "test_type": "quality_control",
  "related_procedures": [],
  "platform_validation": {
    "required": false,
    "evidence_status": "not_executed",
    "real_device_or_at_claimed": false,
    "required_environment_ids": [
      "accessibility-quality-control"
    ],
    "calibration_required": true,
    "evidence_refs": [],
    "limitations": "Quality-control evidence measures method reliability; it is not product accessibility or conformance evidence."
  },
  "library_layer": "composite",
  "effort_guidance": {
    "unit": "minutes",
    "setup": {
      "typical": 30,
      "includes": "Confirm scope and target, prepare safe data, establish the required state and record the tool or environment version."
    },
    "per_target_execution": {
      "typical": 45,
      "includes": "Perform the procedure once for one target in its selected states; this excludes remediation and retesting."
    },
    "evidence_capture": {
      "typical": 20,
      "includes": "Capture, name and cross-reference the minimum reproducible evidence while removing unnecessary personal or secret data."
    },
    "technical_review": {
      "typical": 30,
      "includes": "Check applicability, observations, outcome logic, evidence sufficiency and report wording before issue publication."
    },
    "typical_total_range": {
      "minimum": 125,
      "maximum": 181
    },
    "assumptions": "one frozen review set and one reviewer pair; disagreements, adjudication and reruns are additional effort",
    "variation": "Add a new per-target execution and evidence allowance for each additional target class or state that could produce a different result. Add setup time for each distinct environment.",
    "excludes": "Client coordination, inaccessible-environment recovery, recruitment, legal review, remediation implementation and waiting time unless the procedure states otherwise."
  },
  "worked_examples": [
    {
      "example_id": "A11Y-QC-002-WORKED-001",
      "example_kind": "representative_synthetic_example",
      "context": "Representative synthetic example for Version-frozen calibration set on a fictional, version-frozen product target; no client or participant execution occurred.",
      "inspected": "The tester follows S2 (Collect blinded independent decisions) using the target, required state, tool and repeat rule stated in this procedure.",
      "observed_output": "After the tester completes the Version-frozen calibration set scenario named Collect blinded independent decisions, the recorded observation for the target is inconsistent with the required result: Original decisions are complete, independently produced and immutable.",
      "evidence_captured": [
        "example-evidence-001: target and state record for Each execution agent's independent outcome and evidence for every passed, failed, inapplicable, cantTell and untested calibration fixture.",
        "example-evidence-002: contemporaneous observation from the named method — Agent/fixture IDs, versions, outcomes, evidence, rationale, exception decisions and completion timestamp.",
        "example-evidence-003: comparison note linking the actual observation to the expected result and claim boundary"
      ],
      "tester_reasoning": "The recorded observation does not demonstrate the expected result: Original decisions are complete, independently produced and immutable. The failed decision is limited to this synthetic target and state; unexecuted coverage is not inferred.",
      "decision": "failed",
      "completed_finding": {
        "finding_id": "A11Y-QC-002-example-finding-001",
        "title": "Run the version-frozen execution-team calibration: expected result is not demonstrated",
        "severity": "medium",
        "user_impact": "The represented barrier may affect People affected by assessment decisions whose reliability depends on consistent, evidence-based procedure execution.",
        "evidence_refs": [
          "example-evidence-001",
          "example-evidence-002",
          "example-evidence-003"
        ]
      },
      "remediation_advice": "Change the represented target or process so it demonstrates this result: Original decisions are complete, independently produced and immutable. Then rerun the affected step and every dependent variation.",
      "reviewer_feedback": "The reviewer confirms applicability, environment, actual-versus-expected comparison, evidence references and bounded wording; unsupported product-wide or conformance language is removed.",
      "final_client_report_wording": "Synthetic training finding: the sampled Run the version-frozen execution-team calibration target does not meet the procedure expectation in the recorded state. User impact and broader scope require evidence from the buyer's real execution.",
      "retest_outcome": "Synthetic retest example: after the represented change, repeat the failed step plus material error and recovery states. Close the finding only when new evidence demonstrates the expected result; otherwise retain the non-pass outcome.",
      "claim_boundary": "This is a representative, synthetic worked example for training and calibration. It is not evidence from a client product, participant or real assistive-technology execution."
    }
  ]
}