Implement PMEM-WP-0015 credentialed live pilot with ops-warden routing.

Add credential routing advisories via warden route/access, live pilot evidence
helpers, managed deployment pilot probes, evaluation trend regression gates,
and expanded troubleshooting. Update operator runbook and maturity scorecard.

tests/test_evaluation_scenarios.py

@@ -7,9 +7,11 @@ from phase_memory.contracts import graph_from_markitect
 from phase_memory.evaluation import (
     EVALUATION_REPORT_SCHEMA,
     EVALUATION_TREND_HISTORY_SCHEMA,
+    EVALUATION_TREND_REGRESSION_GATE_SCHEMA,
     EVALUATION_TREND_SCHEMA,
     evaluation_threshold_report,
     evaluation_trend_artifact,
+    evaluation_trend_regression_gate,
     load_evaluation_trend_history,
     write_evaluation_trend_history,
 )
@@ -159,6 +161,37 @@ def test_evaluation_trend_history_persists_without_duplicate_runs(tmp_path) -> N
     assert "policy_denial_count" in loaded["metric_keys"]
 
 
+def test_evaluation_trend_regression_gate_flags_metric_declines() -> None:
+    data = json.loads((FIXTURES / "evaluation-scenarios.json").read_text(encoding="utf-8"))
+    report = evaluation_threshold_report(data)
+    previous = evaluation_trend_artifact(
+        report,
+        run_metadata={"run_id": "previous", "created_at": "2026-05-18T00:00:00+00:00"},
+    )
+    regressed_report = {
+        **report,
+        "metrics": {
+            **report["metrics"],
+            "policy_denial_count": report["metrics"]["policy_denial_count"] - 1,
+        },
+    }
+    latest = evaluation_trend_artifact(
+        regressed_report,
+        previous_report=report,
+        run_metadata={"run_id": "latest", "created_at": "2026-05-19T00:00:00+00:00"},
+    )
+    history = {
+        "schema_version": EVALUATION_TREND_HISTORY_SCHEMA,
+        "artifacts": [previous, latest],
+    }
+
+    gate = evaluation_trend_regression_gate(history)
+
+    assert gate["schema_version"] == EVALUATION_TREND_REGRESSION_GATE_SCHEMA
+    assert gate["valid"] is False
+    assert gate["metric_regressions"]["policy_denial_count"] == -1.0
+
+
 def _activation_plan(response):
     data = response["data"]["activation_plan"]
     return ActivationPlan(