eval.yaml: Hub listing is not a grade. Cite GET /api/gspc. Do not invent a %. Never LLM-judge.

eval.yaml

@@ -1,14 +1,10 @@
-name: GSPC affect
-description: >
-  Frozen GSPC affect measurement bank (Council of AI / csoai).
-  Deterministic grade on frozen splits. Verdict is VALID / INVALID /
-  UNCHECKABLE from an Ed25519 card at https://councilof.ai — not a
-  model-as-judge score. Measurement, never certification. Empty stays
-  empty. Do not wrap this bank in Inspect `model_graded_fact`.
-  Models opt in by pointing Hub Benchmarks at this dataset. Do not
-  open Community Eval PRs without a VALID card.
-evaluation_framework: gspc
-tasks:
-  - id: gspc-affect
-    config: default
-    split: test
+# Hub listing is not a GSPC grade. Cite GET https://councilof.ai/api/gspc
+name: gspc-hub-shell
+cite: https://councilof.ai/api/gspc
+public_count: "22 axis · 15 measured"
+status: coverage
+hub_listing_is_not_a_grade: true
+do_not_invent_percentage: true
+never_llm_judge: true
+never_stamp_measured_from_listing: true
+schema: csoai.gspc-axes/0.5