diff --git a/.scratch/distributed-gguf-runtime/evidence/DGR-025/README.md b/.scratch/distributed-gguf-runtime/evidence/DGR-025/README.md
index 771144c..893c104 100644
--- a/.scratch/distributed-gguf-runtime/evidence/DGR-025/README.md
+++ b/.scratch/distributed-gguf-runtime/evidence/DGR-025/README.md
@@ -72,17 +72,22 @@ explicitly requires fingerprinting the "runtime pin/patch stack".
   committed manifest only; fetching/patching stays with
   `scripts/llama_cpp_dependency.py` (DGR-027).
 - `packages/node/meshnet_node/runtime_recipe.py` — `runtime_version` is now
-  pin-enforced (`_require_pin`) exactly like `tokenizer_revision`; docstring
-  points at the canonical derivation.
+  pin-enforced (`_require_pin`) exactly like `tokenizer_revision`; for the
+  llama.cpp backend it must also match the canonical
+  `llama.cpp@<40-hex>+patchstack.<64-hex>` grammar.
+- `packages/node/meshnet_node/native_backend.py` — the production native
+  identity seam no longer accepts a caller-supplied runtime string. It derives
+  `runtime_version` directly through `load_runtime_pin()` from the committed
+  lock and rejects a non-llama backend at this llama.cpp-specific boundary.
 - `packages/tracker/meshnet_tracker/recipe.py` — the independent tracker
-  implementation applies the same pin rule in `parse_identity`, keeping the two
-  implementations in step.
-- `tests/test_runtime_pin_identity.py` (new, TDD — written first and observed
-  failing) — 17 deterministic tests: the committed manifest derives a
-  deterministic pin whose axis value is a valid recipe pin; a changed patch
-  byte and a reordered stack each change the runtime identity; every manifest
-  disagreement above fails closed; and both node and tracker reject a moving
-  `runtime_version`.
+  implementation applies the same backend-specific grammar before re-deriving
+  the recipe digest, so forged operator labels cannot register or certify.
+- `tests/test_runtime_pin_identity.py` and
+  `tests/test_native_identity_emission.py` — deterministic tests cover lock
+  derivation, production native emission, and node/tracker rejection of the
+  forged values from independent review. Conformance vectors were regenerated
+  through `scripts/gen_recipe_fingerprint_vectors.py` for the tightened wire
+  contract.
 
 ### Backlog-consistency repair (pre-existing damage, honestly recorded)
 
@@ -136,7 +141,7 @@ Marked `DGR-025.passes = true` with `completionNotes`; regenerated
 PYTHONPATH=packages/node:packages/tracker python3 -m pytest -q tests/test_runtime_pin_identity.py
 ```
 ```text
-17 passed in 0.11s
+23 passed in 0.15s
 ```
 
 ```bash
@@ -146,7 +151,7 @@ PYTHONPATH=packages/node:packages/tracker python3 -m pytest -q \
   tests/test_node_admission.py tests/test_node_capability.py tests/test_recipe_benchmark.py
 ```
 ```text
-196 passed, 1 warning in 5.35s
+202 passed, 1 warning in 5.38s
 ```
 
 ```bash
@@ -169,13 +174,12 @@ artifact was touched and nothing was written under `/home`.
 
 ## Limitations
 
-- `runtime_pin.py` proves what the *manifest* pins; it does not prove the
-  running binary was built from that manifest. Binding the built native worker
-  to the pin it reports (e.g. embedding the patched-tree hash at build time and
-  echoing it through the DGR-022 status contract) belongs with the native
-  worker stories (DGR-028+/DGR-031); until then `runtime_version` is exactly as
-  trustworthy as the rest of the declared axes — a claim the tracker digests,
-  with real distributed certification as the trust boundary (unchanged design).
+- The production native identity seam now derives the manifest pin and cannot
+  accept an operator-supplied runtime label. It still cannot attest that the
+  running binary was built from those locked bytes. Embedding the patched-tree
+  hash at build time and echoing it through the DGR-022 status contract belongs
+  with DGR-028+/DGR-031; real distributed certification remains the final trust
+  boundary.
 - The DGR-027-recorded blocker stands: `0002-dense-llama-owned-range-loader.patch`
   does not apply cleanly against the pin (DGR-028). That does not affect this
   story: the identity commits to the patch *bytes as committed*, which is
diff --git a/packages/node/meshnet_node/native_backend.py b/packages/node/meshnet_node/native_backend.py
index 33c6841..9bebe74 100644
--- a/packages/node/meshnet_node/native_backend.py
+++ b/packages/node/meshnet_node/native_backend.py
@@ -12,6 +12,7 @@ from __future__ import annotations
 from dataclasses import dataclass
 
 from .native_protocol import BUNDLE_VERSION, SCHEMA_VERSION, pb
+from .runtime_pin import load_runtime_pin
 from .runtime_recipe import (
     ArtifactIdentity,
     DerivativeBinding,
@@ -72,7 +73,6 @@ class NativeNumericalRecipe:
     kv_layout: str
     architecture_adapter: str
     backend_id: str
-    runtime_version: str
     recipe_id: str
     recipe_version: str
     catalogue_version: str
@@ -95,6 +95,11 @@ def shard_identity_from_native_report(inputs: NativeIdentityInputs) -> ShardIden
     report = inputs.loaded_artifact
     pin = inputs.artifact_pin
     recipe = inputs.numerical_recipe
+    if recipe.backend_id.strip().lower() not in {"llama.cpp", "llama-cpp"}:
+        raise RecipeIdentityError(
+            "native llama.cpp identity requires backend_id 'llama.cpp' or 'llama-cpp'"
+        )
+    runtime_version = load_runtime_pin().runtime_version
     artifact = ArtifactIdentity(
         artifact_id=pin.artifact_id,
         revision=pin.revision,
@@ -115,7 +120,7 @@ def shard_identity_from_native_report(inputs: NativeIdentityInputs) -> ShardIden
             tokenizer_revision=inputs.tokenizer_revision,
             architecture_adapter=recipe.architecture_adapter,
             backend_id=recipe.backend_id,
-            runtime_version=recipe.runtime_version,
+            runtime_version=runtime_version,
             boundary_schema_version=recipe.boundary_schema_version,
             protocol_schema_version=recipe.protocol_schema_version,
             recipe_id=recipe.recipe_id,
diff --git a/packages/node/meshnet_node/runtime_recipe.py b/packages/node/meshnet_node/runtime_recipe.py
index fbb947d..b5e8ce2 100644
--- a/packages/node/meshnet_node/runtime_recipe.py
+++ b/packages/node/meshnet_node/runtime_recipe.py
@@ -125,6 +125,10 @@ _AXIS_MISMATCH: Mapping[str, str] = {
 }
 
 _HEX64 = re.compile(r"^[0-9a-f]{64}$")
+_LLAMA_CPP_RUNTIME_PIN = re.compile(
+    r"^llama\.cpp@[0-9a-f]{40}\+patchstack\.[0-9a-f]{64}$"
+)
+_LLAMA_CPP_BACKEND_IDS = frozenset({"llama.cpp", "llama-cpp"})
 
 # A revision that can move is not a pin. DGR-017 learned this on the artifact;
 # it is just as true of a tokenizer.
@@ -194,6 +198,17 @@ def _require_pin(value: Any, what: str) -> str:
     return text
 
 
+def _require_runtime_pin(value: Any, backend_id: Any) -> str:
+    text = _require_pin(value, "recipe.runtime_version")
+    backend = _require_text(backend_id, "recipe.backend_id").strip().lower()
+    if backend in _LLAMA_CPP_BACKEND_IDS and not _LLAMA_CPP_RUNTIME_PIN.fullmatch(text):
+        raise RecipeIdentityError(
+            "'recipe.runtime_version' for llama.cpp must be "
+            "'llama.cpp@<40-hex commit>+patchstack.<64-hex digest>'"
+        )
+    return text
+
+
 def _as_mapping(value: Any, what: str) -> Mapping[str, Any]:
     if not isinstance(value, Mapping):
         raise RecipeIdentityError(
@@ -405,7 +420,7 @@ class RuntimeRecipe:
             else:
                 _require_text(value, f"recipe.{axis}")
         _require_pin(self.tokenizer_revision, "recipe.tokenizer_revision")
-        _require_pin(self.runtime_version, "recipe.runtime_version")
+        _require_runtime_pin(self.runtime_version, self.backend_id)
         _require_text(self.recipe_id, "recipe.recipe_id")
         _require_text(self.recipe_version, "recipe.recipe_version")
         _require_text(self.catalogue_version, "recipe.catalogue_version")
diff --git a/packages/tracker/meshnet_tracker/recipe.py b/packages/tracker/meshnet_tracker/recipe.py
index 2fe5fbe..50b5cc3 100644
--- a/packages/tracker/meshnet_tracker/recipe.py
+++ b/packages/tracker/meshnet_tracker/recipe.py
@@ -70,6 +70,10 @@ STATUS_CERTIFIED = "certified"
 MIN_CERTIFYING_NODES = 2
 
 _HEX64 = re.compile(r"^[0-9a-f]{64}$")
+_LLAMA_CPP_RUNTIME_PIN = re.compile(
+    r"^llama\.cpp@[0-9a-f]{40}\+patchstack\.[0-9a-f]{64}$"
+)
+_LLAMA_CPP_BACKEND_IDS = frozenset({"llama.cpp", "llama-cpp"})
 _MOVING_REFS = frozenset({"main", "master", "head", "latest", "dev", "trunk"})
 
 
@@ -128,6 +132,17 @@ def _pin(value: Any, what: str) -> str:
     return text
 
 
+def _runtime_pin(value: Any, backend_id: Any) -> str:
+    text = _pin(value, "recipe.runtime_version")
+    backend = _text(backend_id, "recipe.backend_id").strip().lower()
+    if backend in _LLAMA_CPP_BACKEND_IDS and not _LLAMA_CPP_RUNTIME_PIN.fullmatch(text):
+        raise RecipeIdentityError(
+            "'recipe.runtime_version' for llama.cpp must bind a 40-hex commit "
+            "and a 64-hex ordered patch-stack digest"
+        )
+    return text
+
+
 def _mapping(value: Any, what: str) -> Mapping[str, Any]:
     if not isinstance(value, Mapping):
         raise RecipeIdentityError(f"{what!r} must be a JSON object")
@@ -333,7 +348,7 @@ def parse_identity(data: Any) -> PresentedIdentity:
         else:
             axes[axis] = _text(value, f"recipe.{axis}")
     _pin(axes["tokenizer_revision"], "recipe.tokenizer_revision")
-    _pin(axes["runtime_version"], "recipe.runtime_version")
+    _runtime_pin(axes["runtime_version"], axes["backend_id"])
 
     identity = PresentedIdentity(
         artifact_id=_text(artifact.get("artifact_id"), "artifact.artifact_id"),
diff --git a/scripts/gen_recipe_fingerprint_vectors.py b/scripts/gen_recipe_fingerprint_vectors.py
index 702ead3..fcbba17 100644
--- a/scripts/gen_recipe_fingerprint_vectors.py
+++ b/scripts/gen_recipe_fingerprint_vectors.py
@@ -49,7 +49,7 @@ _RECIPE = RuntimeRecipe(
     tokenizer_revision="0123456789abcdef",
     architecture_adapter="llama/range-v1",
     backend_id="llama.cpp",
-    runtime_version="llama.cpp@deadbeef+meshnet.1",
+    runtime_version="llama.cpp@" + "d" * 40 + "+patchstack." + "e" * 64,
     recipe_id="example-gguf",
     recipe_version="1",
     catalogue_version="2026.07.1",
diff --git a/tests/data/recipe_fingerprint_vectors.json b/tests/data/recipe_fingerprint_vectors.json
index 353d15b..04d6a84 100644
--- a/tests/data/recipe_fingerprint_vectors.json
+++ b/tests/data/recipe_fingerprint_vectors.json
@@ -8,9 +8,9 @@
         "model_artifact_digest": "8a0f43d6aa49d77834bdb47bcae9f42c886b7ccfe0ac014932b2a2b38697a47b",
         "recipe_id": "example-gguf",
         "recipe_version": "1",
-        "runtime_recipe_digest": "9b14d70b0835a6428457e4888d453649dd0d2e41fc8ac9d84d232c8c237e68fa"
+        "runtime_recipe_digest": "63001e0efeada5b97f2f3562562dc0fd2d9bd8904dc6b7b99a71d98f3e938bd0"
       },
-      "fingerprint_proto_hex": "0a40386130663433643661613439643737383334626462343762636165396634326338383662376363666530616330313439333262326132623338363937613437621240396231346437306230383335613634323834353765343838386434353336343964643064326534316663386163396438346432333263386332333765363866611a0c6578616d706c652d676775662201312a09323032362e30372e31",
+      "fingerprint_proto_hex": "0a40386130663433643661613439643737383334626462343762636165396634326338383662376363666530616330313439333262326132623338363937613437621240363330303165306566656164613562393766326633353632353632646330666432643962643839303464633662376239396137316439386633653933386264301a0c6578616d706c652d676775662201312a09323032362e30372e31",
       "identity": {
         "artifact": {
           "architecture": "dense-llama",
@@ -26,7 +26,7 @@
           "model_artifact_digest": "8a0f43d6aa49d77834bdb47bcae9f42c886b7ccfe0ac014932b2a2b38697a47b",
           "recipe_id": "example-gguf",
           "recipe_version": "1",
-          "runtime_recipe_digest": "9b14d70b0835a6428457e4888d453649dd0d2e41fc8ac9d84d232c8c237e68fa"
+          "runtime_recipe_digest": "63001e0efeada5b97f2f3562562dc0fd2d9bd8904dc6b7b99a71d98f3e938bd0"
         },
         "recipe": {
           "activation_dtype": "bfloat16",
@@ -40,7 +40,7 @@
           "protocol_schema_version": 1,
           "recipe_id": "example-gguf",
           "recipe_version": "1",
-          "runtime_version": "llama.cpp@deadbeef+meshnet.1",
+          "runtime_version": "llama.cpp@dddddddddddddddddddddddddddddddddddddddd+patchstack.eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee",
           "tokenizer_revision": "0123456789abcdef",
           "weight_quantization": "Q4_K_M"
         },
@@ -58,9 +58,9 @@
         "model_artifact_digest": "8a0f43d6aa49d77834bdb47bcae9f42c886b7ccfe0ac014932b2a2b38697a47b",
         "recipe_id": "example-gguf",
         "recipe_version": "1",
-        "runtime_recipe_digest": "9b14d70b0835a6428457e4888d453649dd0d2e41fc8ac9d84d232c8c237e68fa"
+        "runtime_recipe_digest": "63001e0efeada5b97f2f3562562dc0fd2d9bd8904dc6b7b99a71d98f3e938bd0"
       },
-      "fingerprint_proto_hex": "0a40386130663433643661613439643737383334626462343762636165396634326338383662376363666530616330313439333262326132623338363937613437621240396231346437306230383335613634323834353765343838386434353336343964643064326534316663386163396438346432333263386332333765363866611a0c6578616d706c652d676775662201312a09323032362e30372e31",
+      "fingerprint_proto_hex": "0a40386130663433643661613439643737383334626462343762636165396634326338383662376363666530616330313439333262326132623338363937613437621240363330303165306566656164613562393766326633353632353632646330666432643962643839303464633662376239396137316439386633653933386264301a0c6578616d706c652d676775662201312a09323032362e30372e31",
       "identity": {
         "artifact": {
           "architecture": "dense-llama",
@@ -80,7 +80,7 @@
           "model_artifact_digest": "8a0f43d6aa49d77834bdb47bcae9f42c886b7ccfe0ac014932b2a2b38697a47b",
           "recipe_id": "example-gguf",
           "recipe_version": "1",
-          "runtime_recipe_digest": "9b14d70b0835a6428457e4888d453649dd0d2e41fc8ac9d84d232c8c237e68fa"
+          "runtime_recipe_digest": "63001e0efeada5b97f2f3562562dc0fd2d9bd8904dc6b7b99a71d98f3e938bd0"
         },
         "recipe": {
           "activation_dtype": "bfloat16",
@@ -94,7 +94,7 @@
           "protocol_schema_version": 1,
           "recipe_id": "example-gguf",
           "recipe_version": "1",
-          "runtime_version": "llama.cpp@deadbeef+meshnet.1",
+          "runtime_version": "llama.cpp@dddddddddddddddddddddddddddddddddddddddd+patchstack.eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee",
           "tokenizer_revision": "0123456789abcdef",
           "weight_quantization": "Q4_K_M"
         },
diff --git a/tests/test_native_identity_emission.py b/tests/test_native_identity_emission.py
index 5cc19e9..903c78a 100644
--- a/tests/test_native_identity_emission.py
+++ b/tests/test_native_identity_emission.py
@@ -15,6 +15,7 @@ from meshnet_node.native_backend import (
     shard_identity_from_native_report,
 )
 from meshnet_node.native_protocol import SCHEMA_VERSION, pb
+from meshnet_node.runtime_pin import load_runtime_pin
 from meshnet_node.recipe_manifest import parse_recipe_manifest
 from meshnet_tracker.capability import STATE_UNCERTIFIED, evaluate_report
 
@@ -42,7 +43,6 @@ def _inputs(**changes: object) -> NativeIdentityInputs:
         kv_layout="llama-kv-v1",
         architecture_adapter="dense-llama-v1",
         backend_id="llama-cpp",
-        runtime_version="llama.cpp:e920c523",
         recipe_id="native",
         recipe_version="1",
         catalogue_version="2026.07.1",
@@ -84,6 +84,7 @@ def test_native_identity_uses_loaded_report_not_a_caller_range():
     assert (identity.shard_start, identity.shard_end) == (2, 6)
     assert identity.artifact.architecture == "llama"
     assert identity.artifact.layer_count == 8
+    assert identity.recipe.runtime_version == load_runtime_pin().runtime_version
 
 
 def test_native_identity_requires_an_immutable_pin_and_gguf_range():
diff --git a/tests/test_runtime_pin_identity.py b/tests/test_runtime_pin_identity.py
index 7d8ec75..f66802f 100644
--- a/tests/test_runtime_pin_identity.py
+++ b/tests/test_runtime_pin_identity.py
@@ -42,7 +42,7 @@ def _recipe(**changes: object) -> RuntimeRecipe:
         "tokenizer_revision": "0123456789abcdef",
         "architecture_adapter": "llama/range-v1",
         "backend_id": "llama.cpp",
-        "runtime_version": "llama.cpp@deadbeef+meshnet.1",
+        "runtime_version": "llama.cpp@" + "d" * 40 + "+patchstack." + "e" * 64,
         "recipe_id": "example-gguf",
         "recipe_version": "1",
         "catalogue_version": "2026.07.1",
@@ -276,3 +276,37 @@ def test_tracker_rejects_a_moving_runtime_version():
     doc.pop("fingerprint", None)
     with pytest.raises(TrackerRecipeIdentityError, match="moving reference"):
         parse_identity(doc)
+
+
+@pytest.mark.parametrize(
+    "forged",
+    [
+        "llama.cpp@master+patchstack.not-a-digest",
+        "llama.cpp@e920c523+patchstack.forged",
+        "release-that-operator-typed",
+    ],
+)
+def test_node_recipe_rejects_noncanonical_llama_runtime_pins(forged):
+    with pytest.raises(RecipeIdentityError, match="40-hex commit"):
+        _recipe(runtime_version=forged)
+
+
+@pytest.mark.parametrize(
+    "forged",
+    [
+        "llama.cpp@master+patchstack.not-a-digest",
+        "llama.cpp@e920c523+patchstack.forged",
+        "release-that-operator-typed",
+    ],
+)
+def test_tracker_rejects_noncanonical_llama_runtime_pins(forged):
+    vectors = json.loads(
+        (Path(__file__).parent / "data" / "recipe_fingerprint_vectors.json").read_text(
+            encoding="utf-8"
+        )
+    )
+    doc = json.loads(json.dumps(vectors["vectors"][0]["identity"]))
+    doc["recipe"]["runtime_version"] = forged
+    doc.pop("fingerprint", None)
+    with pytest.raises(TrackerRecipeIdentityError, match="40-hex commit"):
+        parse_identity(doc)
diff --git a/tests/test_runtime_recipe_identity.py b/tests/test_runtime_recipe_identity.py
index 94c216c..fc1abe7 100644
--- a/tests/test_runtime_recipe_identity.py
+++ b/tests/test_runtime_recipe_identity.py
@@ -65,7 +65,7 @@ def _recipe(**changes: object) -> RuntimeRecipe:
         "tokenizer_revision": "0123456789abcdef",
         "architecture_adapter": "llama/range-v1",
         "backend_id": "llama.cpp",
-        "runtime_version": "llama.cpp@deadbeef+meshnet.1",
+        "runtime_version": "llama.cpp@" + "d" * 40 + "+patchstack." + "e" * 64,
         "recipe_id": "example-gguf",
         "recipe_version": "1",
         "catalogue_version": "2026.07.1",
@@ -241,7 +241,7 @@ def test_committed_vectors_cover_a_whole_model_and_a_derivative_shard():
         ("tokenizer_revision", "fedcba9876543210"),
         ("architecture_adapter", "llama/range-v2"),
         ("backend_id", "other-backend"),
-        ("runtime_version", "llama.cpp@other+meshnet.1"),
+        ("runtime_version", "llama.cpp@" + "c" * 40 + "+patchstack." + "b" * 64),
         ("boundary_schema_version", 2),
         ("protocol_schema_version", 2),
     ],