diff --git a/.eamos-core/orchestrator/os/localization/labels.yaml b/.eamos-core/orchestrator/os/localization/labels.yaml
new file mode 100644
index 0000000..04f1dbf
--- /dev/null
+++ b/.eamos-core/orchestrator/os/localization/labels.yaml
@@ -0,0 +1,322 @@
+# Chrome labels (RFC-0001 §7, #61) — every rendered chrome string, per language, as DATA.
+# The single source of truth replacing eight duplicated in-code tables: adding a language is a
+# one-file edit here, never a code change. `en` is the declared fallback. A language is supported
+# only when it defines EVERY key (half-localized is worse than honestly unsupported) — enforced by
+# the labels-valid gate. Namespaces preserve intentional per-deliverable differences (the .docx
+# uses title-case, the sheet says "Source / note"), so values must not be merged across them.
+#
+# core.verify   — the suffix inside the ⟨value — …⟩ grounding marker (RFC-0001 §6)
+# core.redacted — the word inside the ⟦…⟧ egress-redaction mask (RFC-0001 §11)
+version: 1
+
+en:
+  core:
+    verify: "to verify"
+    redacted: "redacted"
+    quiz_stem: "What is the value of"
+    quiz_disc: "What is the stance on"
+  md:
+    target: "target"
+    risk: "Risk"
+    ask: "Ask"
+    decision: "Decision"
+    review: "Verify before the room"
+    objective: "Objective"
+    fill: "source"
+    pro: "pro"
+    con: "con"
+    next_step: "Next step"
+    residual_risk: "Residual risk"
+  quiz_md:
+    title: "Interview quiz"
+    graded: "Graded (citation required)"
+    disc: "Discussion (un-scored)"
+    ans: "Answer"
+    src: "source"
+    verify: "to verify"
+  pptx:
+    target: "target"
+    risk: "Risk"
+    ask: "Ask"
+    pro: "pro"
+    con: "con"
+    review: "Verify before the room"
+    fill: "source"
+    next_step: "Next step"
+    residual_risk: "Residual risk"
+  docx:
+    target: "Target"
+    value: "Value"
+    kpi: "KPI"
+    risk: "Risk"
+    ask: "Ask"
+    pro: "pro"
+    con: "con"
+    review: "Verify before the room"
+    fill: "source"
+    preread: "Pre-read"
+    next_step: "Next step"
+    residual_risk: "Residual risk"
+  svg:
+    decisions: "Decisions required"
+    target: "target"
+    verify: "to verify"
+  xlsx:
+    kpi: "KPI"
+    value: "Value"
+    target: "Target"
+    source: "Source / note"
+    review: "Verify before the room"
+  facilitate:
+    agenda: "Agenda"
+    script: "Facilitation script"
+    min: "min"
+    obj: "Objective"
+    talk: "Talking points"
+    minutes: "Minutes"
+    decisions: "Decisions"
+    actions: "Action items"
+    owner: "owner"
+    due: "due"
+    notes: "Notes"
+    attendees: "Attendees"
+    joins_at: "joins at"
+    input_gathering: "Input gathering"
+    live_classification: "Live classification"
+    targeted_questions: "Targeted questions"
+    structured_synthesis: "Structured synthesis"
+    decision_framing: "Decision framing"
+
+it:
+  core:
+    verify: "da verificare"
+    redacted: "redatto"
+    quiz_stem: "Qual è il valore di"
+    quiz_disc: "Qual è la posizione su"
+  md:
+    target: "target"
+    risk: "Rischio"
+    ask: "Richiesta"
+    decision: "Decisione"
+    review: "Da verificare prima della sala"
+    objective: "Obiettivo"
+    fill: "fonte"
+    pro: "pro"
+    con: "contro"
+    next_step: "Prossimo passo"
+    residual_risk: "Rischio residuo"
+  quiz_md:
+    title: "Quiz d'intervista"
+    graded: "Valutate (citazione richiesta)"
+    disc: "Discussione (non valutate)"
+    ans: "Risposta"
+    src: "fonte"
+    verify: "da verificare"
+  pptx:
+    target: "target"
+    risk: "Rischio"
+    ask: "Richiesta"
+    pro: "pro"
+    con: "contro"
+    review: "Da verificare prima della sala"
+    fill: "fonte"
+    next_step: "Prossimo passo"
+    residual_risk: "Rischio residuo"
+  docx:
+    target: "Target"
+    value: "Valore"
+    kpi: "KPI"
+    risk: "Rischio"
+    ask: "Richiesta"
+    pro: "pro"
+    con: "contro"
+    review: "Da verificare prima della sala"
+    fill: "fonte"
+    preread: "Pre-read"
+    next_step: "Prossimo passo"
+    residual_risk: "Rischio residuo"
+  svg:
+    decisions: "Decisioni richieste"
+    target: "target"
+    verify: "valori da verificare"
+  xlsx:
+    kpi: "KPI"
+    value: "Valore"
+    target: "Target"
+    source: "Fonte / nota"
+    review: "Da verificare prima della sala"
+  facilitate:
+    agenda: "Agenda"
+    script: "Script di facilitazione"
+    min: "min"
+    obj: "Obiettivo"
+    talk: "Punti da toccare"
+    minutes: "Verbale"
+    decisions: "Decisioni"
+    actions: "Azioni"
+    owner: "owner"
+    due: "scadenza"
+    notes: "Note"
+    attendees: "Partecipanti"
+    joins_at: "entra da"
+    input_gathering: "Raccolta input"
+    live_classification: "Classificazione dal vivo"
+    targeted_questions: "Domande mirate"
+    structured_synthesis: "Sintesi strutturata"
+    decision_framing: "Inquadramento decisione"
+
+es:
+  core:
+    verify: "por verificar"
+    redacted: "redactado"
+    quiz_stem: "¿Cuál es el valor de"
+    quiz_disc: "¿Cuál es la posición sobre"
+  md:
+    target: "meta"
+    risk: "Riesgo"
+    ask: "Solicitud"
+    decision: "Decisión"
+    review: "Verificar antes de la reunión"
+    objective: "Objetivo"
+    fill: "fuente"
+    pro: "pro"
+    con: "contra"
+    next_step: "Próximo paso"
+    residual_risk: "Riesgo residual"
+  quiz_md:
+    title: "Cuestionario de entrevista"
+    graded: "Evaluadas (cita requerida)"
+    disc: "Discusión (sin puntuar)"
+    ans: "Respuesta"
+    src: "fuente"
+    verify: "por verificar"
+  pptx:
+    target: "meta"
+    risk: "Riesgo"
+    ask: "Solicitud"
+    pro: "pro"
+    con: "contra"
+    review: "Verificar antes de la reunión"
+    fill: "fuente"
+    next_step: "Próximo paso"
+    residual_risk: "Riesgo residual"
+  docx:
+    target: "Meta"
+    value: "Valor"
+    kpi: "KPI"
+    risk: "Riesgo"
+    ask: "Solicitud"
+    pro: "pro"
+    con: "contra"
+    review: "Verificar antes de la reunión"
+    fill: "fuente"
+    preread: "Prelectura"
+    next_step: "Próximo paso"
+    residual_risk: "Riesgo residual"
+  svg:
+    decisions: "Decisiones requeridas"
+    target: "meta"
+    verify: "valores por verificar"
+  xlsx:
+    kpi: "KPI"
+    value: "Valor"
+    target: "Meta"
+    source: "Fuente / nota"
+    review: "Verificar antes de la reunión"
+  facilitate:
+    agenda: "Agenda"
+    script: "Guion de facilitación"
+    min: "min"
+    obj: "Objetivo"
+    talk: "Puntos a tratar"
+    minutes: "Acta"
+    decisions: "Decisiones"
+    actions: "Acciones"
+    owner: "responsable"
+    due: "vencimiento"
+    notes: "Notas"
+    attendees: "Participantes"
+    joins_at: "se une desde"
+    input_gathering: "Recogida de datos"
+    live_classification: "Clasificación en vivo"
+    targeted_questions: "Preguntas dirigidas"
+    structured_synthesis: "Síntesis estructurada"
+    decision_framing: "Encuadre de la decisión"
+
+fr:
+  core:
+    verify: "à vérifier"
+    redacted: "caviardé"
+    quiz_stem: "Quelle est la valeur de"
+    quiz_disc: "Quelle est la position sur"
+  md:
+    target: "cible"
+    risk: "Risque"
+    ask: "Demande"
+    decision: "Décision"
+    review: "À vérifier avant la réunion"
+    objective: "Objectif"
+    fill: "source"
+    pro: "pour"
+    con: "contre"
+    next_step: "Prochaine étape"
+    residual_risk: "Risque résiduel"
+  quiz_md:
+    title: "Quiz d'entretien"
+    graded: "Notées (citation requise)"
+    disc: "Discussion (non notées)"
+    ans: "Réponse"
+    src: "source"
+    verify: "à vérifier"
+  pptx:
+    target: "cible"
+    risk: "Risque"
+    ask: "Demande"
+    pro: "pour"
+    con: "contre"
+    review: "À vérifier avant la réunion"
+    fill: "source"
+    next_step: "Prochaine étape"
+    residual_risk: "Risque résiduel"
+  docx:
+    target: "Cible"
+    value: "Valeur"
+    kpi: "KPI"
+    risk: "Risque"
+    ask: "Demande"
+    pro: "pour"
+    con: "contre"
+    review: "À vérifier avant la réunion"
+    fill: "source"
+    preread: "Pré-lecture"
+    next_step: "Prochaine étape"
+    residual_risk: "Risque résiduel"
+  svg:
+    decisions: "Décisions requises"
+    target: "cible"
+    verify: "valeurs à vérifier"
+  xlsx:
+    kpi: "KPI"
+    value: "Valeur"
+    target: "Cible"
+    source: "Source / note"
+    review: "À vérifier avant la réunion"
+  facilitate:
+    agenda: "Ordre du jour"
+    script: "Script de facilitation"
+    min: "min"
+    obj: "Objectif"
+    talk: "Points à aborder"
+    minutes: "Compte rendu"
+    decisions: "Décisions"
+    actions: "Actions"
+    owner: "responsable"
+    due: "échéance"
+    notes: "Notes"
+    attendees: "Participants"
+    joins_at: "rejoint à"
+    input_gathering: "Recueil des données"
+    live_classification: "Classification en direct"
+    targeted_questions: "Questions ciblées"
+    structured_synthesis: "Synthèse structurée"
+    decision_framing: "Cadrage de la décision"
diff --git a/.eamos-core/tools/eamos_lint.py b/.eamos-core/tools/eamos_lint.py
index 7d6735c..681e648 100644
--- a/.eamos-core/tools/eamos_lint.py
+++ b/.eamos-core/tools/eamos_lint.py
@@ -28,6 +28,8 @@
   classification-valid       — a problem-type classification (if present) uses taxonomy-allowed values.
   questions-valid            — the adaptive question tree + routing stay consistent with the taxonomy.
   topology-valid             — a topology diagram's edges/kinds/patterns are declared and in-vocabulary.
+  labels-valid               — every language in os/localization/labels.yaml defines every chrome key
+                               the en fallback defines (a language is supported when ALL chrome exists).
 """
 
 import os
@@ -36,6 +38,7 @@
 TOOLS = os.path.dirname(os.path.abspath(__file__))
 sys.path.insert(0, TOOLS)
 import _cli      # noqa: E402  (utf8_stdio, #54)
+import labels    # noqa: E402  (chrome labels as data, #61)
 import render  # noqa: E402  (reuses the loader + build_deck_ir)
 import yamlmini  # noqa: E402
 
@@ -275,6 +278,32 @@ def gate_topology_valid(manifest):
     return findings
 
 
+def gate_labels_valid():
+    """Chrome-label completeness (#61): every language in os/localization/labels.yaml defines every
+    key the `en` fallback defines, and nothing more — a language is supported when ALL chrome exists
+    (half-localized is worse than honestly unsupported), and a key unknown to the fallback is drift.
+    Structural + decidable, manifest-independent; vacuous if the file is absent."""
+    findings = []
+    data = labels.load()
+    ref = data.get(labels.FALLBACK) or {}
+    if not ref:
+        return findings
+    for lang, tables in sorted(data.items()):
+        if lang == labels.FALLBACK or not isinstance(tables, dict):
+            continue                                     # skip the reference itself + scalars (version)
+        for ns, keys in sorted(ref.items()):
+            have = tables.get(ns) or {}
+            for k in sorted(keys):
+                if k not in have:
+                    findings.append(("labels-valid", f"language '{lang}' is missing label '{ns}.{k}'"))
+        for ns, keys in sorted(tables.items()):
+            for k in sorted(keys or {}):
+                if k not in (ref.get(ns) or {}):
+                    findings.append(("labels-valid",
+                                     f"label '{ns}.{k}' in '{lang}' is not defined by the '{labels.FALLBACK}' fallback"))
+    return findings
+
+
 def gate_registry_params(manifest):
     """Every deliverable's params are within the registry's declared enums (RFC-0002 §4)."""
     findings = []
@@ -365,6 +394,7 @@ def run_all(manifest, archetype, deck_ir, acc, ledger):
     findings += gate_classification_valid(manifest)
     findings += gate_questions_valid()
     findings += gate_topology_valid(manifest)
+    findings += gate_labels_valid()
     findings += gate_registry_params(manifest)
     findings += gate_rubric(manifest, deck_ir, {g for g, _ in findings})
     findings += gate_confidentiality(manifest, {g for g, _ in findings})
diff --git a/.eamos-core/tools/emit_docx.py b/.eamos-core/tools/emit_docx.py
index 9b8bdf5..6b4ae65 100644
--- a/.eamos-core/tools/emit_docx.py
+++ b/.eamos-core/tools/emit_docx.py
@@ -16,25 +16,15 @@
 import os
 import sys
 
-import _cli  # utf8_stdio (#54)
+import _cli    # utf8_stdio (#54)
+import labels  # chrome labels as data (#61)
 
 AMBER = (0xB8, 0x6B, 0x00)
 MUTED = (0x70, 0x70, 0x70)
 
-LABELS = {
-    "it": {"target": "Target", "value": "Valore", "kpi": "KPI", "risk": "Rischio", "ask": "Richiesta",
-           "pro": "pro", "con": "contro", "review": "Da verificare prima della sala",
-           "fill": "fonte", "preread": "Pre-read",
-           "next_step": "Prossimo passo", "residual_risk": "Rischio residuo"},
-    "en": {"target": "Target", "value": "Value", "kpi": "KPI", "risk": "Risk", "ask": "Ask",
-           "pro": "pro", "con": "con", "review": "Verify before the room",
-           "fill": "source", "preread": "Pre-read",
-           "next_step": "Next step", "residual_risk": "Residual risk"},
-}
-
 
 def _lab(lang, key):
-    return LABELS.get(lang, LABELS["en"]).get(key, LABELS["en"][key])
+    return labels.lab(lang, "docx", key)   # chrome labels as data (#61)
 
 
 def _run(paragraph, text, bold=False, color=None):
diff --git a/.eamos-core/tools/emit_md.py b/.eamos-core/tools/emit_md.py
index 7e25414..c3d7946 100644
--- a/.eamos-core/tools/emit_md.py
+++ b/.eamos-core/tools/emit_md.py
@@ -12,21 +12,13 @@
 import os
 import sys
 
-import _cli  # utf8_stdio (#54)
-
-# Template chrome in the deck's output language (the section labels, not the content).
-LABELS = {
-    "it": {"target": "target", "risk": "Rischio", "ask": "Richiesta", "decision": "Decisione",
-           "review": "Da verificare prima della sala", "objective": "Obiettivo", "fill": "fonte",
-           "pro": "pro", "con": "contro", "next_step": "Prossimo passo", "residual_risk": "Rischio residuo"},
-    "en": {"target": "target", "risk": "Risk", "ask": "Ask", "decision": "Decision",
-           "review": "Verify before the room", "objective": "Objective", "fill": "source",
-           "pro": "pro", "con": "con", "next_step": "Next step", "residual_risk": "Residual risk"},
-}
+import _cli    # utf8_stdio (#54)
+import labels  # chrome labels as data (#61)
 
 
 def _lab(lang, key):
-    return LABELS.get(lang, LABELS["en"]).get(key, LABELS["en"][key])
+    # Template chrome in the deck's output language (the section labels, not the content).
+    return labels.lab(lang, "md", key)
 
 
 def render_md(ir):
@@ -87,17 +79,11 @@ def render_md(ir):
     return "\n".join(out).rstrip() + "\n"
 
 
-_QUIZ_L = {
-    "it": {"title": "Quiz d'intervista", "graded": "Valutate (citazione richiesta)",
-           "disc": "Discussione (non valutate)", "ans": "Risposta", "src": "fonte", "verify": "da verificare"},
-    "en": {"title": "Interview quiz", "graded": "Graded (citation required)",
-           "disc": "Discussion (un-scored)", "ans": "Answer", "src": "source", "verify": "to verify"},
-}
 
 
 def render_quiz_md(ir):
     lang = ir.get("output_lang", "en")
-    t = _QUIZ_L.get(lang, _QUIZ_L["en"])
+    t = labels.table(lang, "quiz_md")
     out = [f"# {t['title']} — {ir.get('title', '')}", ""]
     graded = [q for q in ir.get("questions", []) if q.get("kind") == "graded"]
     disc = [q for q in ir.get("questions", []) if q.get("kind") == "discussion"]
diff --git a/.eamos-core/tools/emit_pptx.py b/.eamos-core/tools/emit_pptx.py
index 620a42e..bc87114 100644
--- a/.eamos-core/tools/emit_pptx.py
+++ b/.eamos-core/tools/emit_pptx.py
@@ -16,25 +16,17 @@
 import os
 import sys
 
-import _cli  # utf8_stdio (#54)
+import _cli    # utf8_stdio (#54)
+import labels  # chrome labels as data (#61)
 
 NAVY = (0x1E, 0x27, 0x61)      # title
 BODY = (0x2B, 0x2B, 0x2B)      # body text
 AMBER = (0xB8, 0x6B, 0x00)     # assumed / to-verify
 MUTED = (0x70, 0x70, 0x70)     # captions
 
-LABELS = {
-    "it": {"target": "target", "risk": "Rischio", "ask": "Richiesta", "pro": "pro", "con": "contro",
-           "review": "Da verificare prima della sala", "fill": "fonte",
-           "next_step": "Prossimo passo", "residual_risk": "Rischio residuo"},
-    "en": {"target": "target", "risk": "Risk", "ask": "Ask", "pro": "pro", "con": "con",
-           "review": "Verify before the room", "fill": "source",
-           "next_step": "Next step", "residual_risk": "Residual risk"},
-}
-
 
 def _lab(lang, key):
-    return LABELS.get(lang, LABELS["en"]).get(key, LABELS["en"][key])
+    return labels.lab(lang, "pptx", key)   # chrome labels as data (#61)
 
 
 def _line(block, lang):
diff --git a/.eamos-core/tools/emit_svg.py b/.eamos-core/tools/emit_svg.py
index 7c94fe6..227457e 100644
--- a/.eamos-core/tools/emit_svg.py
+++ b/.eamos-core/tools/emit_svg.py
@@ -18,7 +18,8 @@
 import os
 import sys
 
-import _cli  # utf8_stdio (#54)
+import _cli    # utf8_stdio (#54)
+import labels  # chrome labels as data (#61)
 
 # Per visual_style palette (background, accent, body, amber, muted, card). Unknown styles fall back
 # to professional. Layout is shared; only the palette/theme changes (RFC-0002 §11-1).
@@ -26,14 +27,10 @@
     "professional": ("#FFFFFF", "#1E2761", "#2B2B2B", "#B86B00", "#707070", "#EEF1F7"),
     "scientific":   ("#FFFFFF", "#0B3D5C", "#22303A", "#B8860B", "#5A6B73", "#EAF1F5"),
 }
-LABELS = {
-    "it": {"decisions": "Decisioni richieste", "target": "target", "verify": "valori da verificare"},
-    "en": {"decisions": "Decisions required", "target": "target", "verify": "to verify"},
-}
 
 
 def _lab(lang, key):
-    return LABELS.get(lang, LABELS["en"]).get(key, LABELS["en"][key])
+    return labels.lab(lang, "svg", key)    # chrome labels as data (#61)
 
 
 def esc(s):
diff --git a/.eamos-core/tools/emit_xlsx.py b/.eamos-core/tools/emit_xlsx.py
index 9434a80..34f0dfb 100644
--- a/.eamos-core/tools/emit_xlsx.py
+++ b/.eamos-core/tools/emit_xlsx.py
@@ -16,18 +16,11 @@
 import os
 import sys
 
-import _cli  # utf8_stdio (#54)
-
-LABELS = {
-    "it": {"kpi": "KPI", "value": "Valore", "target": "Target", "source": "Fonte / nota",
-           "review": "Da verificare prima della sala"},
-    "en": {"kpi": "KPI", "value": "Value", "target": "Target", "source": "Source / note",
-           "review": "Verify before the room"},
-}
-
+import _cli    # utf8_stdio (#54)
+import labels  # chrome labels as data (#61)
 
 def _lab(lang, key):
-    return LABELS.get(lang, LABELS["en"]).get(key, LABELS["en"][key])
+    return labels.lab(lang, "xlsx", key)   # chrome labels as data (#61)
 
 
 def build_xlsx(ir, out_path):
diff --git a/.eamos-core/tools/facilitate.py b/.eamos-core/tools/facilitate.py
index 3bb17f5..bc004c3 100644
--- a/.eamos-core/tools/facilitate.py
+++ b/.eamos-core/tools/facilitate.py
@@ -25,30 +25,13 @@
 TOOLS = os.path.dirname(os.path.abspath(__file__))
 sys.path.insert(0, TOOLS)
 import _cli      # noqa: E402  (utf8_stdio, #54)
+import labels    # noqa: E402  (chrome labels as data, #61)
 import yamlmini  # noqa: E402
 import render    # noqa: E402
 import series    # noqa: E402
 
-LABELS = {
-    "it": {"agenda": "Agenda", "script": "Script di facilitazione", "min": "min", "obj": "Obiettivo",
-           "talk": "Punti da toccare", "minutes": "Verbale", "decisions": "Decisioni",
-           "actions": "Azioni", "owner": "owner", "due": "scadenza", "notes": "Note",
-           "attendees": "Partecipanti", "joins_at": "entra da",
-           "input_gathering": "Raccolta input", "live_classification": "Classificazione dal vivo",
-           "targeted_questions": "Domande mirate", "structured_synthesis": "Sintesi strutturata",
-           "decision_framing": "Inquadramento decisione"},
-    "en": {"agenda": "Agenda", "script": "Facilitation script", "min": "min", "obj": "Objective",
-           "talk": "Talking points", "minutes": "Minutes", "decisions": "Decisions",
-           "actions": "Action items", "owner": "owner", "due": "due", "notes": "Notes",
-           "attendees": "Attendees", "joins_at": "joins at",
-           "input_gathering": "Input gathering", "live_classification": "Live classification",
-           "targeted_questions": "Targeted questions", "structured_synthesis": "Structured synthesis",
-           "decision_framing": "Decision framing"},
-}
-
-
 def _lab(lang, k):
-    return LABELS.get(lang, LABELS["en"]).get(k, LABELS["en"][k])
+    return labels.lab(lang, "facilitate", k)   # chrome labels as data (#61)
 
 
 def _load(path, what="manifest"):
diff --git a/.eamos-core/tools/labels.py b/.eamos-core/tools/labels.py
new file mode 100644
index 0000000..ed794f3
--- /dev/null
+++ b/.eamos-core/tools/labels.py
@@ -0,0 +1,47 @@
+#!/usr/bin/env python3
+"""Shared chrome-label loader (#61) — the three-tier language model's chrome as data.
+
+os/localization/labels.yaml is the single source of truth for every rendered chrome string: the
+grounding-marker suffix, the redaction mask, the deck / sheet / SVG / quiz / facilitation labels.
+It replaces eight duplicated in-code tables that had already drifted (RFC-0001 §7): adding a
+language is a one-file data edit, and the labels-valid gate enforces that every language defines
+every key — a language is supported when ALL chrome exists, half-localized is worse than honestly
+unsupported.
+
+Namespaces preserve intentional per-deliverable differences (the .docx uses title-case "Target",
+the sheet says "Source / note"): core, md, quiz_md, pptx, docx, svg, xlsx, facilitate.
+"""
+
+import os
+
+TOOLS = os.path.dirname(os.path.abspath(__file__))
+PATH = os.path.join(os.path.dirname(TOOLS), "orchestrator", "os", "localization", "labels.yaml")
+FALLBACK = "en"
+
+_cache = None
+
+
+def load():
+    """The full labels table (cached), or {} if the file is absent."""
+    global _cache
+    if _cache is None:
+        if os.path.exists(PATH):
+            import yamlmini
+            with open(PATH, encoding="utf-8") as fh:
+                _cache = yamlmini.load_yaml(fh.read())
+        else:
+            _cache = {}
+    return _cache
+
+
+def table(lang, ns):
+    """The `ns` label table for `lang`; missing keys fall back per-key to English."""
+    data = load()
+    base = (data.get(FALLBACK) or {}).get(ns) or {}
+    cur = (data.get(lang) or {}).get(ns) or {}
+    return {**base, **cur}
+
+
+def lab(lang, ns, key):
+    """One chrome label; '' only if the key is unknown to the fallback too (labels-valid catches it)."""
+    return table(lang, ns).get(key, "")
diff --git a/.eamos-core/tools/render.py b/.eamos-core/tools/render.py
index 9061562..5e0f8bc 100644
--- a/.eamos-core/tools/render.py
+++ b/.eamos-core/tools/render.py
@@ -20,6 +20,7 @@
 TOOLS = os.path.dirname(os.path.abspath(__file__))
 sys.path.insert(0, TOOLS)
 import _cli      # noqa: E402  (utf8_stdio, #54)
+import labels    # noqa: E402  (chrome labels as data, #61)
 import yamlmini  # noqa: E402
 
 CORE = os.path.dirname(TOOLS)
@@ -32,8 +33,6 @@
 ROUTING = os.path.join(CORE, "orchestrator", "os", "intake", "routing.yaml")
 
 BIND_RE = re.compile(r"\{\{\s*([a-z][a-z0-9_.]*)\s*\}\}")
-# The "verify before the room" label for an assumed value, by output language (RFC-0001 §6).
-VERIFY_LABEL = {"it": "da verificare", "en": "to verify", "es": "por verificar", "fr": "à vérifier"}
 # The provenance enum. Anything else — a typo, a missing field, a future value — is treated as
 # assumed (fail closed, #53) and flagged for the grounding gate: only the explicit `sourced`
 # may render plain.
@@ -45,7 +44,8 @@ def _new_acc():
 
 
 def _mark(value, lang):
-    return f"⟨{value} — {VERIFY_LABEL.get(lang, VERIFY_LABEL['en'])}⟩"
+    # The "verify before the room" label for an assumed value (RFC-0001 §6) — chrome as data (#61).
+    return f"⟨{value} — {labels.lab(lang, 'core', 'verify')}⟩"
 
 
 def _cell_assumed(key, cell, acc):
@@ -272,7 +272,7 @@ def apply_redaction(manifest, policy):
     the shared ledger — so EVERY projection is redacted identically (one ledger, RFC-0002 §7)."""
     tags = redact_tags_for(manifest, policy)
     lang = (manifest.get("context", {}) or {}).get("output_lang", "en")
-    mask = "⟦redatto⟧" if lang == "it" else "⟦redacted⟧"
+    mask = f"⟦{labels.lab(lang, 'core', 'redacted')}⟧"
     n = 0
     for cell in (manifest.get("inputs", {}) or {}).values():
         if isinstance(cell, dict) and any(cell.get(t) for t in tags):
@@ -535,10 +535,6 @@ def build_graph_ir(manifest, archetype, altitude=None, function=None):
             "branches": branches, "review_appendix": review_appendix}, acc
 
 
-_QUIZ_STEM = {"it": "Qual è il valore di", "en": "What is the value of"}
-_QUIZ_DISC = {"it": "Qual è la posizione su", "en": "What is the stance on"}
-
-
 def build_quiz_ir(manifest, archetype, altitude=None, function=None):
     """Project an interview quiz (quiz-IR, RFC-0002 §3, §11-3): graded questions cite a ledger
     source (required); discussion questions (from decisions) are un-scored. SAME ledger."""
@@ -548,7 +544,7 @@ def build_quiz_ir(manifest, archetype, altitude=None, function=None):
     ledger = scorecard_ledger(manifest)
     content = manifest.get("content", {}) or {}
     lang = ctx.get("output_lang", "en")
-    stem, disc = _QUIZ_STEM.get(lang, _QUIZ_STEM["en"]), _QUIZ_DISC.get(lang, _QUIZ_DISC["en"])
+    stem, disc = labels.lab(lang, "core", "quiz_stem"), labels.lab(lang, "core", "quiz_disc")
 
     questions = []
     for key, cell in ledger.items():
diff --git a/.eamos-core/tools/tests/test_gates.py b/.eamos-core/tools/tests/test_gates.py
index 44e28df..c05e613 100644
--- a/.eamos-core/tools/tests/test_gates.py
+++ b/.eamos-core/tools/tests/test_gates.py
@@ -175,6 +175,33 @@ def test_sox_makes_grounding_mandatory(self):
         self.assertIn("confidentiality", ids)   # SOX flags the failed mandatory gate
 
 
+class TestLabelsValid(unittest.TestCase):
+    def test_incomplete_language_fails_labels_valid(self):
+        import labels
+        orig = labels._cache
+        labels._cache = {"en": {"core": {"verify": "to verify"}},
+                         "de": {"core": {}}}                    # half-localized language (#61)
+        try:
+            findings = eamos_lint.gate_labels_valid()
+            self.assertIn(("labels-valid", "language 'de' is missing label 'core.verify'"), findings)
+        finally:
+            labels._cache = orig
+
+    def test_key_unknown_to_fallback_fails_labels_valid(self):
+        import labels
+        orig = labels._cache
+        labels._cache = {"en": {"core": {"verify": "to verify"}},
+                         "it": {"core": {"verify": "da verificare", "vrfy": "typo"}}}   # drift (#61)
+        try:
+            findings = eamos_lint.gate_labels_valid()
+            self.assertTrue(any("not defined by the 'en' fallback" in msg for _, msg in findings))
+        finally:
+            labels._cache = orig
+
+    def test_shipped_labels_are_complete(self):
+        self.assertEqual(eamos_lint.gate_labels_valid(), [])    # es/fr gap is closed
+
+
 class TestUserErrors(unittest.TestCase):
     def test_empty_identity_does_not_crash_gates(self):
         arch = render.load_archetype("review")
diff --git a/.eamos-core/tools/tests/test_render.py b/.eamos-core/tools/tests/test_render.py
index eef546e..71040da 100644
--- a/.eamos-core/tools/tests/test_render.py
+++ b/.eamos-core/tools/tests/test_render.py
@@ -93,6 +93,29 @@ def test_invalid_provenance_fails_closed_across_projections(self):
         q = next(q for q in quiz["questions"] if q["kind"] == "graded" and "ARR" in q["q"])
         self.assertTrue(q["assumed"])
 
+    def test_es_deliverable_is_fully_localized(self):
+        # An output_lang the chrome doesn't fully cover used to render MIXED-language (#61):
+        # Spanish marker, English headings. Now every chrome string localizes together.
+        import emit_md
+        m, arch = load("qbr-c-level")
+        m["context"]["output_lang"] = "es"
+        deck, _ = render.build_deck_ir(m, arch)
+        md = emit_md.render_md(deck)
+        self.assertIn("por verificar", md)                     # grounding marker
+        self.assertIn("Riesgo", md)                            # chrome label
+        self.assertIn("Verificar antes de la reunión", md)     # review appendix heading
+        self.assertNotIn("Verify before the room", md)         # no English fallback leaks
+
+    def test_fr_verify_marker_and_chrome(self):
+        import emit_md
+        m, arch = load("qbr-c-level")
+        m["context"]["output_lang"] = "fr"
+        deck, _ = render.build_deck_ir(m, arch)
+        md = emit_md.render_md(deck)
+        self.assertIn("à vérifier", md)
+        self.assertIn("Risque", md)
+        self.assertNotIn("Verify before the room", md)
+
     def test_invalid_provenance_fails_closed_in_topology(self):
         m, arch = load("vendor-prework")
         m["inputs"]["sys.platform"]["provenance"] = "asumed"   # typo must fail closed (#53)