diff --git a/msdmd/SKILL.md b/msdmd/SKILL.md
index 1a6f0be..a891cc1 100644
--- a/msdmd/SKILL.md
+++ b/msdmd/SKILL.md
@@ -146,6 +146,13 @@ A reference implementation in pure stdlib Python lives at
 Both commit to zero non-stdlib dependencies so you can copy them into
 any project.
 
+The Python reference helper owns its `MODULE_BUILD` and field-preservation
+`CONTRACTS` beside the implementation. A consumer that executes this vendored
+helper in its contract audit should reconcile those exact dependency declarations
+and provide a local resolving `CHECKS` witness. The dependency's canonical owner
+and exact source identity remain explicit; local execution does not transfer
+parser ownership to the consumer.
+
 Extension detection refuses ambiguous suffixes rather than sniffing content.
 For example, `.m` can mean Objective-C or MATLAB/Octave and therefore has no
 automatic marker. A caller that already knows the language may still call
diff --git a/msdmd/parsers/universal.py b/msdmd/parsers/universal.py
index e6cb88a..e25feee 100644
--- a/msdmd/parsers/universal.py
+++ b/msdmd/parsers/universal.py
@@ -1,4 +1,27 @@
-# ratios: loc_comments=161:57 imports_exports=4:7 calls_definitions=55:10
+# ratios: loc_comments=161:80 imports_exports=4:7 calls_definitions=55:10
+# === MODULE_BUILD ===
+# id: msdmd_python_reference_parser
+#   module_name: universal
+#   module_kind: instrument
+#   summary: parses canonical line-comment metadata without executing inspected source
+#   owner: The Interdependency skill-lib
+#   public_surface: COMMENT_MARKERS, RATIO_IDS, marker_for, parse_text, parse_file, walk_tree, parse_ratios, parse_ratios_file, ratios_placement
+#   internal_surface: marker and block matching helpers
+#   auth_boundary: none
+#   storage_boundary: read
+#   network_boundary: none
+#   user_data_boundary: read
+#   admin_only: false
+#   tests: tests/test_universal_parser.py
+#   rollout: exact-pinned reference parser propagation
+#   rollback: restore a previously accepted exact parser identity
+# === END MODULE_BUILD ===
+# === CONTRACTS ===
+# id: msdmd_python_parser_preserves_field_names
+#   given: a valid metadata entry uses lowercase snake-case field names containing digits
+#   then: parsed entries retain those field names and string values without executing the inspected source
+#   class: evidence
+# === END CONTRACTS ===
 """Universal msdmd parser — pure stdlib.
 
 Implements the parser contract from ``msdmd/SKILL.md``: extracts every
@@ -253,4 +276,4 @@ def ratios_placement(text: str, marker: str = "#") -> tuple[bool, bool]:
         last_ok = bool(line_re.match(raw.rstrip()))
         break
     return (opening_ok, last_ok)
-# ratios: loc_comments=161:57 imports_exports=4:7 calls_definitions=55:10
\ No newline at end of file
+# ratios: loc_comments=161:80 imports_exports=4:7 calls_definitions=55:10
diff --git a/tests/test_universal_parser.py b/tests/test_universal_parser.py
index 9cf62bd..33b8860 100644
--- a/tests/test_universal_parser.py
+++ b/tests/test_universal_parser.py
@@ -1,3 +1,12 @@
+# === CHECKS ===
+# id: check_msdmd_python_numeric_field_contract
+#   proves: msdmd_python_parser_preserves_field_names
+#   call: self::test_parser_field_contract
+#   requires: python3
+#   timeout: 10
+#   mutates: filesystem
+#   cleanup: tempdir_teardown
+# === END CHECKS ===
 from __future__ import annotations
 
 import re
@@ -18,6 +27,27 @@
 ROOT = Path(__file__).resolve().parents[1]
 
 
+def test_parser_field_contract() -> None:
+    """No-argument CHECKS witness, also executed by the unittest suite."""
+    checks = unittest.TestCase()
+    helper = ROOT / "msdmd/parsers/universal.py"
+    module = parse_file(helper, "MODULE_BUILD")[0]
+    checks.assertEqual("msdmd_python_reference_parser", module["id"])
+    checks.assertIn("COMMENT_MARKERS", module["public_surface"].split(", "))
+    checks.assertEqual("read", module["storage_boundary"])
+    checks.assertEqual("read", module["user_data_boundary"])
+    checks.assertEqual("msdmd_python_parser_preserves_field_names", parse_file(helper, "CONTRACTS")[0]["id"])
+    with tempfile.TemporaryDirectory() as directory:
+        marker = Path(directory) / "executed"
+        source = Path(directory) / "inspected.py"
+        text = "# === NARRATIVE ===\n# id: source_bound_narrative\n#   evidence_sha256: abc123\n# === END NARRATIVE ===\n"
+        source.write_text(text + "from pathlib import Path\n" + f"Path({str(marker)!r}).write_text('executed')\n")
+        expected = [{"id": "source_bound_narrative", "evidence_sha256": "abc123"}]
+        checks.assertEqual(expected, parse_text(text, "NARRATIVE"))
+        checks.assertEqual(expected, parse_file(source, "NARRATIVE"))
+        checks.assertFalse(marker.exists(), "parsing executed inspected source")
+
+
 class UniversalParserTest(unittest.TestCase):
     def test_parse_single_block_with_multiple_entries(self) -> None:
         text = """# === CONTRACTS ===
@@ -58,15 +88,7 @@ def test_parse_all_matching_blocks_not_just_first(self) -> None:
         )
 
     def test_parse_numeric_snake_case_field(self) -> None:
-        text = """# === NARRATIVE ===
-# id: source_bound_narrative
-#   evidence_sha256: abc123
-# === END NARRATIVE ===
-"""
-        self.assertEqual(
-            [{"id": "source_bound_narrative", "evidence_sha256": "abc123"}],
-            parse_text(text, "NARRATIVE"),
-        )
+        test_parser_field_contract()
 
     def test_parse_typescript_comment_marker(self) -> None:
         text = """// === CAPABILITIES ===