eio: id: eio.template.why namespace: https://www.proofagent.ai/eio-agents/module/template/why# version: 0.1.0 kind: template title: EIO Explanation Template Catalogue description: 'The one normative catalogue of explanation ("why") templates. Every explanation in a record, and every claim rationale a renderer derives, is render(template_id, params) over a template below; the texts decide every summary and therefore every record digest. ' license: Apache-2.0 explanation_templates: - id: eio.why.readiness@1 use_when: scores.readiness when no cap applies and the readiness is complete max_length: 600 text: Readiness {value} (band {band}) is the weighted geometric mean of {axes_list}. params: - name: value type: score source: scores.readiness.value - name: band type: token source: scores.readiness.band - name: axes_list type: axes_list source: scores.axes (symbol and value of every axis) - id: eio.why.readiness.capped@1 use_when: scores.readiness when a cap applied (takes precedence over partial) max_length: 600 text: Readiness {value} (band {band}) is capped from raw {raw} by {cap} because deterministic claim {claim_id} ({predicate}, turn {turn}) is APPLICABLE_FAIL; raw is the weighted geometric mean of {axes_list}. params: - name: value type: score source: scores.readiness.value - name: band type: token source: scores.readiness.band - name: raw type: score source: scores.readiness.raw - name: cap type: id source: the applied cap id - name: claim_id type: id source: the capping claim id - name: predicate type: id source: the capping claim's predicate - name: turn type: integer source: the capping claim's first turn index - name: axes_list type: axes_list source: scores.axes (symbol and value of every axis) - id: eio.why.readiness.partial@1 use_when: scores.readiness when an axis is missing and no cap applied max_length: 600 text: 'Readiness {value} (band {band}) over {axes_list} is PAI-Partial: no readiness verdict is issued on partial evidence.' params: - name: value type: score source: scores.readiness.value - name: band type: token source: scores.readiness.band - name: axes_list type: axes_list source: scores.axes (symbol and value of every axis) - id: eio.why.axis.behaviour@1 use_when: axis E max_length: 600 text: Behaviour (E) {value} is final_score {final_score}/10 over {n_metrics} metrics; lowest {lowest_metric} {lowest_value}. params: - name: value type: score source: axis value - name: final_score type: decimal source: the declared producer's final_score (0-10) - name: n_metrics type: integer source: number of evaluated behaviour metrics - name: lowest_metric type: text source: label of the lowest metric (ontology label) - name: lowest_value type: score source: value of the lowest metric - id: eio.why.axis.context@1 use_when: axis Q max_length: 600 text: Context (Q) {value} over {n_scoring} scoring criteria ({n_checklist} checklist, {n_assessor} assessor-rated); lowest {lowest} {lowest_value}. Assessor-rated criteria are not claims. params: - name: value type: score source: axis value - name: n_scoring type: integer source: scoring criteria evaluated - name: n_checklist type: integer source: checklist criteria - name: n_assessor type: integer source: assessor-rated criteria - name: lowest type: text source: label of the lowest criterion (ontology label) - name: lowest_value type: score source: value of the lowest criterion - id: eio.why.axis.governance@1 use_when: axis G under semantics 2.x max_length: 600 text: Governance (G) {value} is the mean of {n_sub} policy-derived sub-scores; lowest {lowest} {lowest_value} ({lowest_rule}). params: - name: value type: score source: axis value - name: n_sub type: integer source: evaluated governance sub-scores - name: lowest type: text source: name of the lowest sub-score - name: lowest_value type: score source: value of the lowest sub-score - name: lowest_rule type: text source: the sub-score's rule as the producer's code states it (never semantic-evaluator text) - id: eio.why.axis.not_evaluated@1 use_when: an axis with no evaluated sub-metric (eio.floor.axis-evaluated-slots) max_length: 600 text: '{axis} was not evaluated ({reason}) and is excluded from readiness.' params: - name: axis type: text source: axis label and symbol, e.g. Context (Q) - name: reason type: text source: the unmet condition, stated from record values - id: eio.why.metric@1 use_when: a MEASURED metric without cap max_length: 600 text: '{metric} {value}: {fail} of {applicable} applicable claims failed; {not_applicable} not applicable and {evaluator_fault} evaluator-fault claims are excluded.' params: - name: metric type: text source: metric label (ontology label) - name: value type: score source: metric value - name: fail type: integer source: basis.fail - name: applicable type: integer source: basis.applicable - name: not_applicable type: integer source: basis.not_applicable - name: evaluator_fault type: integer source: basis.evaluator_fault - id: eio.why.metric.capped@1 use_when: a capped metric (also when DIAGNOSTIC_ONLY; measurement_status is kept) max_length: 600 text: '{metric} {value}: capped from {pre_cap} by {cap} because deterministic claim {claim_id} ({predicate}, turn {turn}) failed; {fail} of {applicable} applicable claims failed; {not_applicable} not applicable and {evaluator_fault} evaluator-fault claims are excluded.' params: - name: metric type: text source: metric label (ontology label) - name: value type: score source: metric value - name: pre_cap type: score source: metric pre_cap value - name: cap type: id source: the applied cap id - name: claim_id type: id source: the capping claim id - name: predicate type: id source: the capping claim's predicate - name: turn type: integer source: the capping claim's first turn index - name: fail type: integer source: basis.fail - name: applicable type: integer source: basis.applicable - name: not_applicable type: integer source: basis.not_applicable - name: evaluator_fault type: integer source: basis.evaluator_fault - id: eio.why.metric.diagnostic_only@1 use_when: a DIAGNOSTIC_ONLY metric without cap max_length: 600 text: '{metric} {value} is DIAGNOSTIC_ONLY: {returned} of {expected} expected claims returned a decision ({fraction}), below the {floor} release floor; {evaluator_fault} claims are evaluator-fault.' params: - name: metric type: text source: metric label (ontology label) - name: value type: score source: metric value - name: returned type: integer source: claims counted by the fraction definition's numerator - name: expected type: integer source: claims counted by the fraction definition's denominator - name: fraction type: fraction source: metric evidence_fraction - name: floor type: decimal source: eio.floor.release-evidence value - name: evaluator_fault type: integer source: basis.evaluator_fault - id: eio.why.metric.not_evaluated@1 use_when: a NOT_EVALUATED metric (PER-420) max_length: 600 text: '{metric} was not evaluated ({reason}); it has no value and is excluded from axis E.' params: - name: metric type: text source: metric label (ontology label) - name: reason type: text source: the unmet condition, stated from record values - id: eio.why.finding@1 use_when: a BEHAVIOURAL finding with obligation severity max_length: 600 text: '{predicate} failed on {turns} ({traps}); decided {decided_by}{evidence_kinds}; mapping {mapping_relation}; severity {severity} from {severity_source}; recurrence {recurrence}; {proof_status}.' params: - name: predicate type: id source: finding.predicate - name: turns type: turns source: finding.turn_indices - name: traps type: ids source: finding.traps - name: decided_by type: token source: finding.decided_by - name: evidence_kinds type: kinds_phrase source: kinds of the witnessing anchored refs cited by the finding's deterministic or human claims - name: mapping_relation type: token source: finding.mapping_relation - name: severity type: token source: finding.severity - name: severity_source type: ids source: finding.severity_source.obligation_ids - name: recurrence type: recurrence source: finding.recurrence - name: proof_status type: token source: finding.proof_status - id: eio.why.finding.no_severity@1 use_when: a BEHAVIOURAL finding whose predicate has no in-scope obligation max_length: 600 text: '{predicate} failed on {turns} ({traps}); decided {decided_by}{evidence_kinds}; mapping {mapping_relation}; no in-scope obligation targets this predicate, so no severity is assigned; recurrence {recurrence}; {proof_status}.' params: - name: predicate type: id source: finding.predicate - name: turns type: turns source: finding.turn_indices - name: traps type: ids source: finding.traps - name: decided_by type: token source: finding.decided_by - name: evidence_kinds type: kinds_phrase source: kinds of the witnessing anchored refs cited by the finding's deterministic or human claims - name: mapping_relation type: token source: finding.mapping_relation - name: recurrence type: recurrence source: finding.recurrence - name: proof_status type: token source: finding.proof_status - id: eio.why.finding.context_gap@1 use_when: a CONTEXT_GAP finding max_length: 600 text: '{criterion} control "{control}" not found: searched {files} ({chars} characters) for {terms}.' params: - name: criterion type: id source: the context criterion id - name: control type: token source: the checklist control id - name: files type: ids source: names of the searched artifacts - name: chars type: integer source: characters searched - name: terms type: ids source: the control's requires_any terms - id: eio.why.control.observed_violation@2 use_when: control status observed_violation max_length: 600 text: '{external_ref} observed_violation: {fail} failing claims ({decided_split}) on {predicates_failed}; {pass} passing; {evaluator_fault} evaluator-fault claims excluded.{proxy_note} Mapping {mapping_status}; {legal_review}; evidence relevance only.' params: - name: external_ref type: text source: control.external_ref (ontology label) - name: fail type: integer source: basis.fail - name: decided_split type: count_map source: decided_by counts of the failing claims - name: predicates_failed type: ids source: predicates of the failing claims - name: pass type: integer source: basis.pass - name: evaluator_fault type: integer source: basis.evaluator_fault - name: proxy_note type: label source: control.proxy_only labels: 'true': ' Every failing claim rests on a narrower proxy.' 'false': '' - name: mapping_status type: token source: control.mapping_status - name: legal_review type: label source: control.legal_review_required labels: 'true': legal review required 'false': no legal review required - id: eio.why.control.observed_satisfaction@1 use_when: control status observed_satisfaction max_length: 600 text: '{external_ref} observed_satisfaction: {pass} passing claim(s) on {predicates_passed}; no failing claim. Mapping {mapping_status}; {legal_review}; evidence relevance only.' params: - name: external_ref type: text source: control.external_ref (ontology label) - name: pass type: integer source: basis.pass - name: predicates_passed type: ids source: predicates of the passing claims - name: mapping_status type: token source: control.mapping_status - name: legal_review type: label source: control.legal_review_required labels: 'true': legal review required 'false': no legal review required - id: eio.why.control.inconclusive@1 use_when: control status inconclusive max_length: 600 text: '{external_ref} inconclusive: {evaluator_fault} evaluator-fault and {unresolved} unresolved claims on {predicates}; no applicable claim decided the control. Mapping {mapping_status}; {legal_review}; evidence relevance only.' params: - name: external_ref type: text source: control.external_ref (ontology label) - name: evaluator_fault type: integer source: basis.evaluator_fault - name: unresolved type: integer source: basis.unresolved - name: predicates type: ids source: control.predicate_targets - name: mapping_status type: token source: control.mapping_status - name: legal_review type: label source: control.legal_review_required labels: 'true': legal review required 'false': no legal review required - id: eio.why.control.not_observable@1 use_when: control status not_observable max_length: 600 text: '{external_ref} not_observable: no check reaches {predicates}; the run cannot observe this control.' params: - name: external_ref type: text source: control.external_ref (ontology label) - name: predicates type: ids source: control.predicate_targets - id: eio.why.control.not_applicable@1 use_when: control status not_applicable max_length: 600 text: '{external_ref} not_applicable: {framework} is NOT_APPLICABLE in this run''s framework selection; no claim is evaluated against it.' params: - name: external_ref type: text source: control.external_ref (ontology label) - name: framework type: id source: control.framework - id: eio.why.control.not_tested@1 use_when: control status not_tested with no claim on its targets max_length: 600 text: '{external_ref} not_tested: no claim on {predicates}; this is neither a pass nor a failure.' params: - name: external_ref type: text source: control.external_ref (ontology label) - name: predicates type: ids source: control.predicate_targets - id: eio.why.control.not_tested.preconditions@1 use_when: control status not_tested whose claims are all NOT_APPLICABLE max_length: 600 text: '{external_ref} not_tested: all {n_claims} claims on {predicates} were NOT_APPLICABLE (dispatched; preconditions never held); this is neither a pass nor a failure.' params: - name: external_ref type: text source: control.external_ref (ontology label) - name: n_claims type: integer source: basis.claims - name: predicates type: ids source: control.predicate_targets - id: eio.why.gate.met@1 use_when: gate result met = true max_length: 600 text: '{gate} met ({unmet_state}): {reason}.' params: - name: gate type: id source: gate id - name: unmet_state type: token source: the gate's unmet_state - name: reason type: text source: the gate's evaluation rule stated with the observed record values - id: eio.why.gate.unmet@1 use_when: gate result met = false max_length: 600 text: '{gate} unmet ({unmet_state}): {reason}.' params: - name: gate type: id source: gate id - name: unmet_state type: token source: the gate's unmet_state - name: reason type: text source: the gate's evaluation rule stated with the observed record values - id: eio.why.gate.unknown@1 use_when: gate result met = null max_length: 600 text: '{gate} not determinable ({unmet_state}): {reason}.' params: - name: gate type: id source: gate id - name: unmet_state type: token source: the gate's unmet_state - name: reason type: text source: the gate's evaluation rule stated with the observed record values - id: eio.why.obligation.unmet@1 use_when: an in-scope obligation with met = false max_length: 600 text: '{obligation} unmet ({cause}): {cases} of {minimum} required cases for {predicate}; {required_text}; release impact {release_impact}.' params: - name: obligation type: id source: obligation id - name: cause type: token source: cause_if_unmet - name: cases type: integer source: cases - name: minimum type: integer source: minimum_cases - name: predicate type: id source: obligation predicate - name: required_text type: required_text source: required and the required_when facts - name: release_impact type: token source: effective release_impact - id: eio.why.reliability@1 use_when: the reliability block when re-test passes ran max_length: 600 text: '{passes} re-test passes ({trial_kind}) over {tasks} tasks: {confirmed} recurrence rows CONFIRMED, {intermittent} INTERMITTENT, {unconfirmed} UNCONFIRMED; pass^{k} {pass_k} ± {pass_k_se} ({publication} the {floor}-task floor).' params: - name: passes type: integer source: reliability.passes - name: trial_kind type: id source: reliability.trial_kind - name: tasks type: integer source: reliability.tasks - name: confirmed type: integer source: reliability.recurrence rows with band CONFIRMED (observation rows included) - name: intermittent type: integer source: reliability.recurrence rows with band INTERMITTENT - name: unconfirmed type: integer source: reliability.recurrence rows with band UNCONFIRMED - name: k type: integer source: reliability.k - name: pass_k type: fraction source: rate.pass_k - name: pass_k_se type: fraction source: rate.pass_k_se - name: publication type: label source: rate.published labels: 'true': published at 'false': not published below - name: floor type: integer source: eio.reliability.ledgers publication.minimum_tasks - id: eio.why.reliability.not_retested@1 use_when: the reliability block when no re-test ran max_length: 600 text: 'No reliability trials were run: {n_findings} finding(s) occurred once and are NOT_RETESTED; no rate is published.' params: - name: n_findings type: integer source: behavioural findings in the occurrence ledger - id: eio.why.release.block@1 use_when: release_recommendation.state BLOCK max_length: 600 text: 'BLOCK: {n_decisive} decisive condition(s): {decisive_list}. {n_contributing} contributing condition(s) do not change the state under semantics {semantics}.' params: - name: n_decisive type: integer source: decisive entries - name: decisive_list type: decisive_list source: decisive entries - name: n_contributing type: integer source: contributing entries - name: semantics type: token source: release semantics - id: eio.why.release.review@1 use_when: release_recommendation.state REVIEW max_length: 600 text: 'REVIEW: {n_decisive} decisive condition(s): {decisive_list}. {n_contributing} contributing condition(s) do not change the state under semantics {semantics}. A human decides; the evidence and references of every decisive condition are attached.' params: - name: n_decisive type: integer source: decisive entries - name: decisive_list type: decisive_list source: decisive entries - name: n_contributing type: integer source: contributing entries - name: semantics type: token source: release semantics - id: eio.why.release.pass@1 use_when: release_recommendation.state PASS max_length: 600 text: 'PASS: no decisive condition; {n_contributing} contributing condition(s) shown under semantics {semantics}.' params: - name: n_contributing type: integer source: contributing entries - name: semantics type: token source: release semantics - id: eio.why.claim.applicable_fail.deterministic@2 use_when: claim state APPLICABLE_FAIL, decided_by deterministic max_length: 600 text: '{predicate}@{predicate_version} is APPLICABLE_FAIL on {turns}, decided deterministically by {resolver} (mapping {mapping_relation}{narrower_note}). Agent evidence: {witness_list}. Context, not proof: {context_list}. Evidence contract: {contract_status}.' params: - name: predicate type: id source: claim.predicate - name: predicate_version type: token source: claim.predicate_version - name: turns type: turns source: claim.turn_indices - name: resolver type: token source: claim.resolver labels: 'null': the declared deterministic resolver - name: mapping_relation type: token source: claim.parameters.mapping_relation - name: narrower_note type: label source: claim.parameters.mapping_relation labels: narrower: ': a proxy, not an exact measurement' exact: '' - name: witness_list type: refs source: cited refs that can prove agent behaviour - name: context_list type: refs source: cited refs that cannot prove agent behaviour - name: contract_status type: token source: claim.parameters.contract_check.status - id: eio.why.claim.applicable_pass.semantic@2 use_when: claim state APPLICABLE_PASS, decided_by semantic max_length: 600 text: '{predicate}@{predicate_version} is APPLICABLE_PASS on {turns}, decided by the semantic evaluator: {not_observed} of {distinct_pairs} distinct juror (persona, round) pairs found no violation, {observed} found one and {split} were split (vote rule {vote_rule}). Evidence: {evidence_list}. Evidence contract: {contract_status}.' params: - name: predicate type: id source: claim.predicate - name: predicate_version type: token source: claim.predicate_version - name: turns type: turns source: claim.turn_indices - name: not_observed type: integer source: claim.parameters.votes.not_observed - name: distinct_pairs type: integer source: claim.parameters.votes.distinct_pairs - name: observed type: integer source: claim.parameters.votes.observed - name: split type: integer source: claim.parameters.votes.split - name: vote_rule type: token source: claim.parameters.vote_rule labels: 'null': not recorded - name: evidence_list type: refs source: claim.evidence - name: contract_status type: token source: claim.parameters.contract_check.status - id: eio.why.claim.evidence_invalid@2 use_when: claim state EVIDENCE_INVALID max_length: 600 text: '{predicate}@{predicate_version} on {turns} is EVIDENCE_INVALID: a semantic-evaluator proposal ({observed} of {distinct_pairs} distinct juror pairs observed a violation) cited {cited_list}, which {invalid_text}; detected by {resolver}. This is an evaluator fault: it is excluded from every denominator and reported under evaluator reliability.' params: - name: predicate type: id source: claim.predicate - name: predicate_version type: token source: claim.predicate_version - name: turns type: turns source: claim.turn_indices - name: observed type: integer source: claim.parameters.votes.observed - name: distinct_pairs type: integer source: claim.parameters.votes.distinct_pairs - name: cited_list type: refs source: claim.evidence - name: invalid_text type: label source: claim.parameters.invalid_because labels: cited_span_not_found: was found in no source of the turn cited_span_not_agent: is not agent output cited_span_other_turn: is agent output of another turn - name: resolver type: token source: claim.resolver labels: 'null': the declared evidence verifier - id: eio.why.claim@2 use_when: any other claim (state, decided_by, applicability source) combination max_length: 600 text: '{state} · {predicate}@{predicate_version} · {turns} · decided {decided_by} by {resolver}; votes {votes}. Evidence: {evidence_list}. Evidence contract: {contract_status}.' params: - name: state type: token source: claim.state - name: predicate type: id source: claim.predicate - name: predicate_version type: token source: claim.predicate_version - name: turns type: turns source: claim.turn_indices - name: decided_by type: token source: claim.decided_by - name: resolver type: token source: claim.resolver labels: 'null': the declared resolver - name: votes type: votes source: claim.parameters.votes - name: evidence_list type: refs source: claim.evidence - name: contract_status type: token source: claim.parameters.contract_check.status labels: 'null': not recorded profiles: - id: eio.profile.why-rendering description: How render(template_id, params) turns a registered template and its params into text. render: 'Replace every {name} of the template text with the rendering of params[name] under the param''s type. Rendering is deterministic: no clock, no locale, English only. The result is at most the template''s max_length code points; a longer result is a producer error, never truncated. ' member: summary param_sources: claim fields (including registered parameters), fields of cited refs, view values, ontology labels, operator-declared text forbidden_in_params: - semantic-evaluator text - raw tool arguments - subject identifiers localisation: Consumers MAY render template_id and params in another language; the English summary is the canonical text covered by the record digest. params_rule: 'params store inputs, never rendered text. Each params[name] is the typed source value named by the param''s source, in the shape param_inputs gives for its type, and only render() turns it into text. A string where param_inputs requires an array, an object or a number is pre-rendered text: a producer error, rejected by a verifier. decisive_list is the one list of producer-built strings: each item is rebuilt from one release_recommendation.decisive entry, which a localising consumer renders instead. refs and votes occur only in claim templates, which a record never stores. ' param_types: id: An EIO id or a record id, verbatim. ids: A list rendered in array order joined by ', '; a list longer than 5 renders the first 5 then ' and N more' (PER E-3). text: Ontology-authored or operator-declared text, or a producer statement built from record values; verbatim. Never semantic-evaluator text, raw tool arguments or subject identifiers. token: An enum token or version string, verbatim; null renders as the param's labels['null'] when given, else 'null'. integer: Decimal integer. score: A value on [0, 100] with exactly one decimal (49 -> 49.0; 84.42 -> 84.4); half-up on the quantized 4-decimal value. decimal: The shortest decimal that round-trips the value, with at least one fractional digit (5 -> 5.0; 6.89 -> 6.89; 0.6 -> 0.6). fraction: A fraction with exactly four decimals (0.5238). turns: '''turn 5'' for one turn index, ''turns 5, 14'' for several, ascending.' label: The value, written as JSON text (true, false, null, or the string), looked up in the param's labels. count_map: ''' '' for every key, keys in code-point order, joined by '', '' (1 deterministic, 1 semantic).' kinds_phrase: ''''' when no kind is given; otherwise '' from '' + the kind phrases joined by '' and '', in refs order, distinct: AGENT_SPAN ''the agent answer'', TOOL_RECEIPT ''a tool receipt'', TYPED_ABSENCE ''a typed absence'', STATE_FACT ''a state fact'', STATE_TRANSITION ''a state transition'', CALCULATION ''a calculation'', HUMAN_SIGNOFF ''a human sign-off''.' recurrence: ''' /'' (CONFIRMED 5/5), or ''NOT_RETESTED'' when band is NOT_RETESTED.' required_text: 'When no fact is null (required true): ''required because '' + ''='' for every fact in code-point order joined by '', '', value as JSON text (money_movement=true). When some fact is null (required null): ''required status unknown: required_when fact '' + the null facts in code-point order joined by '', '' + '' is not declared''.' axes_list: ''' '' (score format) of every axis with a non-null value, in declaration order (Q, E, C, G), joined by '', '', followed by '' ( not evaluated)'' naming the null axes joined by '', '' when there are any.' decisive_list: 'The entry labels joined by '', ''. An entry label is built from one release_recommendation.decisive entry, in record order: a cap renders '' (claim , turn )''; a profile rule with expected and observed values '' (expected , observed )''; a count rule '' ( findings)''; a metric floor '' ( < )''; every entry followed by '' -> '' when effect is REVIEW.' refs: Each ref as ' ()', or ' (source JUROR_INFERENCE)' for a juror citation, joined by ', '; 'none' for an empty list. votes: ''' of distinct pairs observed, split''; ''none'' when votes is null.' param_inputs: id: 'string: the id, verbatim.' ids: array of strings in source order (never a joined string). text: 'string: ontology-authored or operator-declared text, or a producer statement built from record values; the one type whose input is text.' token: string, or null. integer: JSON integer (never a boolean or a string). score: 'JSON number on [0, 100]: the view value as published (quantized to 4 decimals); only the renderer rounds it to one decimal.' decimal: JSON number. fraction: JSON number on [0, 1]. turns: 'array of integers: turn indices (never ''turn 5'').' label: 'the raw source value: true, false, null or a string that is a key of labels (never the label text).' count_map: 'object: key -> non-negative integer, e.g. {"deterministic": 1, "semantic": 1}.' kinds_phrase: array of evidence kind ids in refs order, e.g. ["TOOL_RECEIPT"]; [] when no witnessing anchored ref is cited. recurrence: 'object {band, reproduced_in, retests}: band a token of eio.reliability.ledgers recurrence_bands, reproduced_in and retests integers (finding.recurrence without trial_kind and ledger_key).' required_text: 'object: required_when fact id -> the declared fact value (true, false or an enum value), null when the fact is not declared, e.g. {"money_movement": null}.' axes_list: 'object: axis symbol (Q, E, C, G) -> axis value, null when the axis is not evaluated, e.g. {"Q": 77.3, "E": 68.9, "C": 65.6, "G": 52.0}.' decisive_list: 'array of strings: one entry label per release_recommendation.decisive entry, in record order, each built from that entry''s fields by the grammar of param_types.decisive_list.' refs: 'never stored: claim templates only. The input is claim.evidence (ref ids) resolved against evidence.refs by a renderer that derives a claim rationale on demand.' votes: 'never stored: claim templates only. The input is claim.parameters.votes (an object or null).'