codex/synthesis.py

@@ -102,6 +102,13 @@ class Lead:
     # Derived: claim-level grounding decisions (empty for conjectures).
     claims: list[Claim] = field(default_factory=list)
 
+    def __post_init__(self) -> None:
+        if not self.provenance:
+            raise ValueError(
+                f"Lead '{self.id}' must carry at least one Provenance entry. "
+                "For gap leads with no chunks, use a topic-marker Provenance."
+            )
+
 
 # ---------------------------------------------------------------------------
 # Protocols
@@ -432,7 +439,7 @@ def find_gaps(
                         f"The retrieval layer returned no chunks for topic '{topic}'.\n"
                         "No grounded statement is possible; this is the absence itself."
                     ),
-                    provenance=[],
+                    provenance=[Provenance(bibkey="(none)", locator=f"topic:{topic}")],
                     confidence=1.0,
                     status="unverified",
                     suggested_validation=(
@@ -797,12 +804,7 @@ def write_leads(leads: list[Lead], leads_dir: str) -> None:
 
 
 def _update_index(root: Path) -> None:
-    """Re-build ``root/INDEX.md`` from the on-disk files.
-
-    Append-only semantics: previously listed ids that no longer have a
-    file are still recorded under "Removed" so the index never silently
-    loses provenance.
-    """
+    """Re-build ``root/INDEX.md`` from the on-disk files (full rebuild, not append-only)."""
     grounded_dir = root / "grounded"
     conj_dir = root / "conjectures"
     grounded_files = sorted(grounded_dir.glob("*.md")) if grounded_dir.exists() else []

tests/synthesis/test_model.py

@@ -78,6 +78,19 @@ def test_lead_missing_provenance_field_raises() -> None:
         )
 
 
+def test_lead_empty_provenance_raises() -> None:
+    """Constructing Lead with empty provenance list raises ValueError."""
+    with pytest.raises(ValueError, match="at least one Provenance"):
+        Lead(
+            id="L-0004",
+            kind="connection",
+            title="x",
+            body="y",
+            provenance=[],
+            confidence=0.5,
+        )
+
+
 def test_lead_id_format_helper() -> None:
     """The internal _make_lead_id helper produces zero-padded L-XXXX strings."""
     from codex.synthesis import _make_lead_id