Parse only explicit manifest paths

scripts/audit_data_operations.py

@@ -143,6 +143,23 @@ def _iter_strings(value: Any) -> Iterable[str]:
             yield from _iter_strings(nested)
 
 
+def _iter_manifest_path_values(value: Any, parent_key: str = "") -> Iterable[str]:
+    path_tokens = ("path", "file", "artifact", "manifest", "output")
+    if isinstance(value, dict):
+        for key, nested in value.items():
+            normalized_key = str(key).strip().lower()
+            if isinstance(nested, str) and any(token in normalized_key for token in path_tokens):
+                yield nested
+            else:
+                yield from _iter_manifest_path_values(nested, normalized_key)
+    elif isinstance(value, (list, tuple)):
+        for nested in value:
+            if isinstance(nested, str) and any(token in parent_key for token in path_tokens):
+                yield nested
+            else:
+                yield from _iter_manifest_path_values(nested, parent_key)
+
+
 def normalize_storage_reference(value: str | None, storage_root: Path) -> Path | None:
     if not value:
         return None
@@ -182,7 +199,7 @@ def expand_manifest_references(references: set[Path], storage_root: Path) -> set
             payload = json.loads(manifest.read_text(encoding="utf-8"))
         except (OSError, UnicodeDecodeError, json.JSONDecodeError):
             continue
-        for value in _iter_strings(payload):
+        for value in _iter_manifest_path_values(payload):
             normalized = normalize_storage_reference(value, root)
             if normalized is None and "://" not in value:
                 try: