fix: drop the sklearn declaration — no loadable artifact exists here

library_name: sklearn plus sklearn/joblib tags told the Hub to present this as a loadable scikit-learn model, but the repo ships no joblib, npz or safetensors artifact, and get_kernel is UNAVAILABLE so the kernel path does not resolve either. Removes the declaration and those tags, and adds a NOT A LOADABLE MODEL banner. The surrogate rules and tests are real; the checkpoint is not.

README.md

@@ -1,15 +1,23 @@
 ---
 license: apache-2.0
-library_name: sklearn
 tags:
-  - sklearn
-  - joblib
-  - surrogate
-  - recipe-conformance
-  - szl-holdings
-  - doctrine-v11
+- surrogate
+- recipe-conformance
+- szl-holdings
+- doctrine-v11
+- not-a-checkpoint
 ---
 
+> **NOT A LOADABLE MODEL.** This repository contains no weight artifact — no
+> `joblib`, no `npz`, no `safetensors`. It previously declared
+> `library_name: sklearn` with `sklearn` and `joblib` tags, which told the Hub to
+> present it as a loadable scikit-learn model; nothing here can be loaded that
+> way. `get_kernel` is **UNAVAILABLE** (returns `False`), so the kernel path does
+> not resolve either. Both the sklearn declaration and the kernel path have been
+> removed from the metadata rather than left to imply a capability that is
+> absent. The surrogate rules and tests in this repo are real; the checkpoint is
+> not.
+
 <!-- SZL-KERNEL-OPERATIONAL:START -->
 ## Operational (MEASURED laptop-Blackwell)