chore(repo): make the release gate reproducible and clean product URLs

PyJHora absence is now partial, tests write research manifests to tmp, the registry allows experimental_variant, and product links point at the Gitea repo. Early logs move to docs/history.

Co-authored-by: Cursor <cursoragent@cursor.com>

scripts/capability_evidence_pool.py

@@ -30,6 +30,7 @@ def build_capability_evidence_pool_summary(registry: dict[str, Any] | None = Non
     prediction_counts = Counter()
     primary_entries: list[str] = []
     audit_only_entries: list[str] = []
+    comparison_only_entries: list[str] = []
     alias_entries: list[str] = []
 
     for tech_id, tech in techniques.items():
@@ -50,6 +51,8 @@ def build_capability_evidence_pool_summary(registry: dict[str, Any] | None = Non
             primary_entries.append(tech_id)
         elif evidence_role == "audit_only":
             audit_only_entries.append(tech_id)
+        elif evidence_role == "comparison_only":
+            comparison_only_entries.append(tech_id)
         elif evidence_role == "alias":
             alias_entries.append(tech_id)
 
@@ -67,12 +70,14 @@ def build_capability_evidence_pool_summary(registry: dict[str, Any] | None = Non
         "prediction_verification_counts": dict(sorted(prediction_counts.items())),
         "primary_entries": sorted(primary_entries),
         "audit_only_entries": sorted(audit_only_entries),
+        "comparison_only_entries": sorted(comparison_only_entries),
         "alias_entries": sorted(alias_entries),
         "conclusion_policy": {
             "primary_chain_required": True,
             "all_89_entries_must_not_be_flattened_into_conclusions": True,
             "supporting_entries_can_only_raise_or_lower_confidence": True,
             "audit_only_entries_cannot_affect_astrological_conclusions": True,
+            "comparison_only_entries_cannot_affect_astrological_conclusions": True,
             "conflicts_must_downgrade_confidence": True,
         },
     }