diff --git a/docs/REPORT_SCHEMA.md b/docs/REPORT_SCHEMA.md
index 761c7b2..f3e4036 100644
--- a/docs/REPORT_SCHEMA.md
+++ b/docs/REPORT_SCHEMA.md
@@ -224,6 +224,87 @@ Current metrics include:
 - runtime dependency staging grouped by bundled plugin when OpenClaw emits
   `runtimeDeps.stage` spans with `pluginId` attributes
 
+## Health And Readiness
+
+Records keep existing compatibility fields such as `timeToListeningMs`,
+`timeToHealthReadyMs`, `readinessClassification`, `healthFailures`, and
+`healthP95Ms`. New readers should use `records[*].measurements.health`:
+
+```json
+{
+  "schemaVersion": "kova.health.v1",
+  "readiness": {
+    "phaseId": "cold-start",
+    "listeningReadyAtMs": 2536,
+    "healthReadyAtMs": 3005,
+    "classification": "ready",
+    "severity": "pass",
+    "reason": "gateway became healthy within the readiness threshold",
+    "thresholdMs": 30000,
+    "deadlineMs": 120000,
+    "attempts": 4
+  },
+  "startupSamples": {
+    "scope": "startup-sample",
+    "count": 4,
+    "okCount": 1,
+    "failureCount": 3,
+    "p95Ms": 120,
+    "maxMs": 120,
+    "slowestPhaseId": "cold-start"
+  },
+  "postReadySamples": {
+    "scope": "post-ready",
+    "count": 9,
+    "okCount": 9,
+    "failureCount": 0,
+    "p95Ms": 469,
+    "maxMs": 652,
+    "slowestPhaseId": "api-latency"
+  },
+  "unknownSamples": {
+    "scope": "unknown",
+    "count": 0,
+    "okCount": 0,
+    "failureCount": 0,
+    "p95Ms": null,
+    "maxMs": null,
+    "slowestPhaseId": null
+  },
+  "final": {
+    "scope": "final",
+    "gatewayState": "running",
+    "ok": true,
+    "healthOk": true,
+    "failureCount": 0,
+    "p95Ms": 90,
+    "maxMs": 90,
+    "slowestPhaseId": "final"
+  },
+  "slowestSample": {
+    "scope": "post-ready",
+    "phaseId": "api-latency",
+    "durationMs": 652
+  }
+}
+```
+
+Scenario phases declare `healthScope` so the evaluator does not infer meaning
+from phase ids. Allowed values are `readiness`, `startup-sample`, `post-ready`,
+`final`, and `none`. Old or externally produced reports without phase scope are
+treated as `unknown` when summarized for compatibility.
+
+Compatibility derivation:
+
+- `timeToListeningMs`: `measurements.health.readiness.listeningReadyAtMs`
+- `timeToHealthReadyMs`: `measurements.health.readiness.healthReadyAtMs`
+- `readinessClassification`: `measurements.health.readiness.classification`
+- `healthFailures`: startup + post-ready + unknown + final health failures
+- `healthP95Ms`: max startup/post-ready p95, falling back to old aggregate p95
+  for old reports
+- `startupHealthP95Ms`: `measurements.health.startupSamples.p95Ms`
+- `postReadyHealthP95Ms`: `measurements.health.postReadySamples.p95Ms`
+
 Role-specific thresholds can fail a scenario separately from total process-tree
 thresholds. For example, a report can show that `gateway` exceeded memory while
 `package-manager` stayed normal, or that `package-manager` spiked during local
@@ -276,8 +357,8 @@ Aggregate metric fields include:
 - `samples`
 
 Current aggregate metrics include startup readiness, TCP listening, RSS, CPU,
-event-loop delay, agent turn latency, health p95, and runtime dependency
-staging.
+event-loop delay, agent turn latency, compatibility health p95, startup health
+p95, post-ready health p95, and runtime dependency staging.
 
 Baseline stores use schema `kova.baselines.v1`. Baseline read/write requires
 `--execute` so stored evidence comes from real OpenClaw runs, not dry-run plans.
diff --git a/metrics/known.json b/metrics/known.json
index 92fa8b9..58ef29e 100644
--- a/metrics/known.json
+++ b/metrics/known.json
@@ -30,6 +30,7 @@
     "diagnosticPresent",
     "doctorFixMs",
     "eventLoopMaxMs",
+    "finalHealthFailures",
     "gatewayReadyHardTimeoutMs",
     "gatewayReadyMs",
     "gatewayResponsive",
@@ -64,6 +65,8 @@
     "pluginIndexPresent",
     "pluginInstallMs",
     "pluginLoadFailures",
+    "postReadyHealthFailures",
+    "postReadyHealthP95Ms",
     "pluginUpdateDryRunMs",
     "pluginsListMs",
     "preProviderDominanceRatio",
@@ -89,6 +92,8 @@
     "statusAfterFailureMs",
     "statusAfterModelsMs",
     "statusMs",
+    "startupHealthFailures",
+    "startupHealthP95Ms",
     "syncFsStallDetected",
     "tuiSmokeMs",
     "upgradeMs",
diff --git a/scenarios/agent-auth-missing.json b/scenarios/agent-auth-missing.json
index ef6e6f2..e2d7a88 100644
--- a/scenarios/agent-auth-missing.json
+++ b/scenarios/agent-auth-missing.json
@@ -42,7 +42,8 @@
         "runtime binding",
         "startup readiness",
         "no Kova auth setup phase"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "missing-auth-agent-turn",
@@ -57,7 +58,8 @@
         "no provider request",
         "process leak snapshot",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-auth-failure-health",
@@ -72,7 +74,8 @@
         "auth failure logs",
         "plugin errors",
         "memory after auth failure"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/agent-cold-warm-message.json b/scenarios/agent-cold-warm-message.json
index e5be32d..ca3c666 100644
--- a/scenarios/agent-cold-warm-message.json
+++ b/scenarios/agent-cold-warm-message.json
@@ -43,7 +43,8 @@
         "gateway port",
         "runtime binding",
         "env created without service"
-      ]
+      ],
+      "healthScope": "none"
     },
     {
       "id": "cold-agent-turn",
@@ -58,7 +59,8 @@
         "mock provider request timing",
         "gateway health after cold turn",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "warm-agent-turn",
@@ -73,7 +75,8 @@
         "mock provider request timing",
         "cold/warm delta",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-agent-health",
@@ -86,7 +89,8 @@
         "env status",
         "plugin errors",
         "memory after agent turns"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/agent-gateway-rpc-turn.json b/scenarios/agent-gateway-rpc-turn.json
index 3c2121a..662b41a 100644
--- a/scenarios/agent-gateway-rpc-turn.json
+++ b/scenarios/agent-gateway-rpc-turn.json
@@ -37,7 +37,8 @@
         "gateway port",
         "runtime binding",
         "env created without service"
-      ]
+      ],
+      "healthScope": "none"
     },
     {
       "id": "gateway-start",
@@ -51,7 +52,8 @@
         "gateway service installed",
         "gateway service started",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "gateway-agent-turn",
@@ -66,7 +68,8 @@
         "mock provider request timing",
         "gateway health after turn",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-agent-health",
@@ -81,7 +84,8 @@
         "provider logs",
         "plugin errors",
         "memory after agent turn"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/agent-long-session.json b/scenarios/agent-long-session.json
index 2bea2bd..c1d8196 100644
--- a/scenarios/agent-long-session.json
+++ b/scenarios/agent-long-session.json
@@ -46,7 +46,8 @@
         "gateway port",
         "runtime binding",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "cold-session-turn",
@@ -60,7 +61,8 @@
         "assistant text",
         "provider request timing",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "warm-session-turn",
@@ -74,7 +76,8 @@
         "assistant text",
         "provider request timing",
         "cold/warm delta"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "session-turn-3",
@@ -88,7 +91,8 @@
         "assistant text",
         "provider request timing",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "session-turn-4",
@@ -102,7 +106,8 @@
         "assistant text",
         "provider request timing",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "session-turn-5",
@@ -116,7 +121,8 @@
         "assistant text",
         "provider request timing",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "session-turn-6",
@@ -130,7 +136,8 @@
         "assistant text",
         "provider request timing",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-session-health",
@@ -146,7 +153,8 @@
         "plugin errors",
         "memory after repeated turns",
         "process leak summary"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/agent-network-offline.json b/scenarios/agent-network-offline.json
index afdbbfd..528ec36 100644
--- a/scenarios/agent-network-offline.json
+++ b/scenarios/agent-network-offline.json
@@ -36,7 +36,8 @@
         "gateway port",
         "runtime binding",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "network-offline-turn",
@@ -49,7 +50,8 @@
         "bounded network failure",
         "gateway status after failure",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-network-health",
@@ -64,7 +66,8 @@
         "network/provider failure logs",
         "plugin errors",
         "memory after network failure"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/agent-provider-concurrent.json b/scenarios/agent-provider-concurrent.json
index 79ddef9..8fe2c45 100644
--- a/scenarios/agent-provider-concurrent.json
+++ b/scenarios/agent-provider-concurrent.json
@@ -47,7 +47,8 @@
         "gateway port",
         "runtime binding",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "concurrent-provider-turns",
@@ -63,7 +64,8 @@
         "pre-provider timing",
         "role resource samples",
         "process leak snapshot"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-concurrency-health",
@@ -78,7 +80,8 @@
         "provider logs",
         "plugin errors",
         "memory after concurrent turns"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/agent-provider-malformed.json b/scenarios/agent-provider-malformed.json
index 3cbc72e..6d89a41 100644
--- a/scenarios/agent-provider-malformed.json
+++ b/scenarios/agent-provider-malformed.json
@@ -38,7 +38,8 @@
         "gateway port",
         "runtime binding",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "malformed-provider-turn",
@@ -53,7 +54,8 @@
         "malformed provider evidence",
         "gateway remains supervised",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-failure-health",
@@ -68,7 +70,8 @@
         "provider logs",
         "plugin errors",
         "memory after malformed response"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/agent-provider-recovery.json b/scenarios/agent-provider-recovery.json
index 367f4da..87661fd 100644
--- a/scenarios/agent-provider-recovery.json
+++ b/scenarios/agent-provider-recovery.json
@@ -42,7 +42,8 @@
         "gateway port",
         "runtime binding",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "transient-provider-failure-turn",
@@ -57,7 +58,8 @@
         "provider 200 recovery evidence",
         "gateway remains supervised",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "recovery-provider-turn",
@@ -71,7 +73,8 @@
         "provider recovery timing",
         "gateway remains healthy",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-failure-health",
@@ -86,7 +89,8 @@
         "provider logs",
         "plugin errors",
         "memory after recovery"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/agent-provider-slow.json b/scenarios/agent-provider-slow.json
index f4c04b3..2ee9eed 100644
--- a/scenarios/agent-provider-slow.json
+++ b/scenarios/agent-provider-slow.json
@@ -44,7 +44,8 @@
         "gateway port",
         "runtime binding",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "slow-provider-turn",
@@ -58,7 +59,8 @@
         "provider delay timing",
         "pre-provider timing",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-failure-health",
@@ -73,7 +75,8 @@
         "provider logs",
         "plugin errors",
         "memory after provider delay"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/agent-provider-streaming-stall.json b/scenarios/agent-provider-streaming-stall.json
index c70a1f2..c6cd8a7 100644
--- a/scenarios/agent-provider-streaming-stall.json
+++ b/scenarios/agent-provider-streaming-stall.json
@@ -41,7 +41,8 @@
         "gateway port",
         "runtime binding",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "streaming-stall-provider-turn",
@@ -57,7 +58,8 @@
         "process leak snapshot",
         "gateway remains supervised",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-failure-health",
@@ -72,7 +74,8 @@
         "provider logs",
         "plugin errors",
         "memory after streaming stall"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/agent-provider-timeout.json b/scenarios/agent-provider-timeout.json
index 26c30c9..14a49f1 100644
--- a/scenarios/agent-provider-timeout.json
+++ b/scenarios/agent-provider-timeout.json
@@ -39,7 +39,8 @@
         "gateway port",
         "runtime binding",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "timeout-provider-turn",
@@ -54,7 +55,8 @@
         "provider timeout/abort timing",
         "gateway remains supervised",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-failure-health",
@@ -69,7 +71,8 @@
         "provider logs",
         "plugin errors",
         "memory after timeout"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/browser-automation-smoke.json b/scenarios/browser-automation-smoke.json
index 582adf0..1e48620 100644
--- a/scenarios/browser-automation-smoke.json
+++ b/scenarios/browser-automation-smoke.json
@@ -39,7 +39,8 @@
         "gateway status",
         "gateway port",
         "readiness classification"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "browser-smoke",
@@ -54,7 +55,8 @@
         "opened tab count",
         "snapshot timing",
         "browser stop timing"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-browser-health",
@@ -68,7 +70,8 @@
         "status after browser automation",
         "browser plugin errors",
         "gateway errors"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/bundled-plugin-startup.json b/scenarios/bundled-plugin-startup.json
index 715069a..2ebe8a2 100644
--- a/scenarios/bundled-plugin-startup.json
+++ b/scenarios/bundled-plugin-startup.json
@@ -30,7 +30,8 @@
         "bundled plugin count",
         "readiness classification",
         "dependency staging"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "inspect",
@@ -46,7 +47,8 @@
         "registry refresh",
         "missing package/module errors",
         "plugin service failures"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "restart",
@@ -61,7 +63,8 @@
         "warm readiness",
         "bundled plugin reload",
         "runtime dependency reuse"
-      ]
+      ],
+      "healthScope": "readiness"
     }
   ],
   "proves": [
diff --git a/scenarios/bundled-runtime-deps.json b/scenarios/bundled-runtime-deps.json
index 7b7863d..374d431 100644
--- a/scenarios/bundled-runtime-deps.json
+++ b/scenarios/bundled-runtime-deps.json
@@ -29,7 +29,8 @@
         "dependency staging duration",
         "installed dependency list",
         "missing dependency errors"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "warm-restart",
@@ -44,7 +45,8 @@
         "warm ready time",
         "dependency staging reuse",
         "missing dependency errors"
-      ]
+      ],
+      "healthScope": "readiness"
     }
   ],
   "proves": [
diff --git a/scenarios/cross-platform-smoke.json b/scenarios/cross-platform-smoke.json
index f60584e..868cdf6 100644
--- a/scenarios/cross-platform-smoke.json
+++ b/scenarios/cross-platform-smoke.json
@@ -37,7 +37,8 @@
         "Node version",
         "runtime version",
         "gateway port"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "core-smoke",
@@ -53,7 +54,8 @@
         "plugin list",
         "filesystem stall logs",
         "health latency"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/dashboard-readiness.json b/scenarios/dashboard-readiness.json
index eb1cfdd..c2e2d98 100644
--- a/scenarios/dashboard-readiness.json
+++ b/scenarios/dashboard-readiness.json
@@ -31,7 +31,8 @@
         "gateway status",
         "gateway port",
         "readiness classification"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "dashboard",
@@ -44,7 +45,8 @@
         "dashboard URL",
         "token handling",
         "command latency"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-dashboard-health",
@@ -58,7 +60,8 @@
         "status after dashboard command",
         "websocket disconnect logs",
         "gateway errors"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/dashboard-session-send-turn-existing-user.json b/scenarios/dashboard-session-send-turn-existing-user.json
index afcfb3a..c737f68 100644
--- a/scenarios/dashboard-session-send-turn-existing-user.json
+++ b/scenarios/dashboard-session-send-turn-existing-user.json
@@ -41,7 +41,8 @@
         "source env",
         "clone root",
         "cloned OpenClaw config"
-      ]
+      ],
+      "healthScope": "none"
     },
     {
       "id": "upgrade",
@@ -55,7 +56,8 @@
         "upgrade JSON",
         "runtime binding",
         "post-upgrade service state"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "gateway-start",
@@ -69,7 +71,8 @@
         "gateway service installed",
         "gateway service started",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "dashboard-session-turn",
@@ -85,7 +88,8 @@
         "provider timing",
         "gateway health after turn",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-dashboard-health",
@@ -101,7 +105,8 @@
         "liveness warnings",
         "plugin errors",
         "memory after dashboard turn"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/dashboard-session-send-turn.json b/scenarios/dashboard-session-send-turn.json
index 362ae6e..1057e39 100644
--- a/scenarios/dashboard-session-send-turn.json
+++ b/scenarios/dashboard-session-send-turn.json
@@ -38,7 +38,8 @@
         "gateway port",
         "runtime binding",
         "env created without service"
-      ]
+      ],
+      "healthScope": "none"
     },
     {
       "id": "gateway-start",
@@ -52,7 +53,8 @@
         "gateway service installed",
         "gateway service started",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "dashboard-session-turn",
@@ -67,7 +69,8 @@
         "mock provider request timing",
         "gateway health after turn",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-dashboard-health",
@@ -82,7 +85,8 @@
         "provider logs",
         "plugin errors",
         "memory after dashboard turn"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/failure-injection.json b/scenarios/failure-injection.json
index 933b2cb..aeecdda 100644
--- a/scenarios/failure-injection.json
+++ b/scenarios/failure-injection.json
@@ -27,7 +27,8 @@
       "evidence": [
         "baseline status",
         "gateway PID"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "diagnostics",
@@ -41,7 +42,8 @@
         "error classification",
         "gateway survival",
         "recovery guidance"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/fresh-install.json b/scenarios/fresh-install.json
index 759d5a6..6c03e68 100644
--- a/scenarios/fresh-install.json
+++ b/scenarios/fresh-install.json
@@ -30,7 +30,8 @@
         "env name",
         "runtime binding",
         "gateway port"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "readiness",
@@ -45,7 +46,8 @@
         "gateway state",
         "gateway PID",
         "health/status result"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "plugins",
@@ -59,7 +61,8 @@
         "plugins list output",
         "plugin update dry-run output",
         "missing dependency log scan"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "models",
@@ -72,7 +75,8 @@
         "models list duration",
         "timeout behavior",
         "gateway health after model list"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "logs",
@@ -85,7 +89,8 @@
         "startup logs",
         "missing dependency errors",
         "plugin metadata scan warnings"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "cleanup",
@@ -96,7 +101,8 @@
       ],
       "evidence": [
         "destroy result"
-      ]
+      ],
+      "healthScope": "none"
     }
   ],
   "proves": [
diff --git a/scenarios/gateway-performance.json b/scenarios/gateway-performance.json
index b076f5d..40f31bb 100644
--- a/scenarios/gateway-performance.json
+++ b/scenarios/gateway-performance.json
@@ -13,9 +13,9 @@
   "thresholds": {
     "coldReadyMs": 30000,
     "warmReadyMs": 15000,
-    "healthP95Ms": 1000,
     "peakRssMb": 900,
-    "eventLoopMaxMs": 500
+    "eventLoopMaxMs": 500,
+    "postReadyHealthP95Ms": 1000
   },
   "phases": [
     {
@@ -32,7 +32,8 @@
         "RSS",
         "CPU",
         "startup logs"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "api-latency",
@@ -47,7 +48,8 @@
         "command durations",
         "health after each command",
         "logs"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "warm-restart",
@@ -61,7 +63,8 @@
         "warm ready time",
         "RSS delta",
         "startup log delta"
-      ]
+      ],
+      "healthScope": "readiness"
     }
   ],
   "proves": [
diff --git a/scenarios/mcp-runtime-start-stop.json b/scenarios/mcp-runtime-start-stop.json
index d74712b..e181f8e 100644
--- a/scenarios/mcp-runtime-start-stop.json
+++ b/scenarios/mcp-runtime-start-stop.json
@@ -37,7 +37,8 @@
         "gateway status",
         "gateway port",
         "readiness classification"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "mcp-bridge",
@@ -51,7 +52,8 @@
         "tools/list timing",
         "tool count",
         "bridge process exit"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-mcp-health",
@@ -65,7 +67,8 @@
         "status after MCP bridge",
         "MCP bridge errors",
         "gateway errors"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/media-understanding-timeout.json b/scenarios/media-understanding-timeout.json
index d42dd60..ce47d91 100644
--- a/scenarios/media-understanding-timeout.json
+++ b/scenarios/media-understanding-timeout.json
@@ -39,7 +39,8 @@
         "gateway port",
         "runtime binding",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "media-timeout",
@@ -53,7 +54,8 @@
         "provider timeout observed",
         "gateway status after timeout",
         "mock provider request log"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-media-health",
@@ -68,7 +70,8 @@
         "provider timeout logs",
         "plugin errors",
         "memory after media timeout"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/official-plugin-install.json b/scenarios/official-plugin-install.json
index dba8c8a..b0a02e7 100644
--- a/scenarios/official-plugin-install.json
+++ b/scenarios/official-plugin-install.json
@@ -36,7 +36,8 @@
       "evidence": [
         "fresh env started",
         "baseline plugin list captured"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "install",
@@ -50,7 +51,8 @@
         "security scanner results",
         "plugins appear in list",
         "registry refresh succeeds"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "restart",
@@ -67,7 +69,8 @@
         "official plugin remains installed",
         "plugin load logs",
         "missing dependency scan"
-      ]
+      ],
+      "healthScope": "readiness"
     }
   ],
   "proves": [
diff --git a/scenarios/openai-compatible-turn.json b/scenarios/openai-compatible-turn.json
index 19c8f33..d5c2b4b 100644
--- a/scenarios/openai-compatible-turn.json
+++ b/scenarios/openai-compatible-turn.json
@@ -37,7 +37,8 @@
         "gateway port",
         "runtime binding",
         "env created without service"
-      ]
+      ],
+      "healthScope": "none"
     },
     {
       "id": "gateway-start",
@@ -51,7 +52,8 @@
         "gateway service installed",
         "gateway service started",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "openai-compatible-turn",
@@ -66,7 +68,8 @@
         "mock provider request timing",
         "gateway health after turn",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-http-health",
@@ -81,7 +84,8 @@
         "provider logs",
         "plugin errors",
         "memory after HTTP turn"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/plugin-bad-manifest.json b/scenarios/plugin-bad-manifest.json
index 060f324..99f9030 100644
--- a/scenarios/plugin-bad-manifest.json
+++ b/scenarios/plugin-bad-manifest.json
@@ -28,7 +28,8 @@
       "evidence": [
         "baseline gateway status",
         "readiness classification"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "reject-invalid-plugin",
@@ -41,7 +42,8 @@
         "install command rejected",
         "validation error",
         "no install record committed"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-failure-health",
@@ -56,7 +58,8 @@
         "gateway status",
         "plugin list",
         "logs after invalid install"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/plugin-external-install.json b/scenarios/plugin-external-install.json
index b3dedd2..9e6e280 100644
--- a/scenarios/plugin-external-install.json
+++ b/scenarios/plugin-external-install.json
@@ -29,7 +29,8 @@
       "evidence": [
         "baseline plugin list",
         "gateway readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "install",
@@ -45,7 +46,8 @@
         "plugin index update",
         "registry refresh",
         "plugin appears in list"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "restart",
@@ -60,7 +62,8 @@
         "restart readiness",
         "plugin load logs",
         "missing dependency scan"
-      ]
+      ],
+      "healthScope": "readiness"
     }
   ],
   "proves": [
diff --git a/scenarios/plugin-lifecycle.json b/scenarios/plugin-lifecycle.json
index 2d9990b..2853cf8 100644
--- a/scenarios/plugin-lifecycle.json
+++ b/scenarios/plugin-lifecycle.json
@@ -28,7 +28,8 @@
         "plugin list",
         "update dry-run",
         "runtime dependency errors"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "restart",
@@ -43,7 +44,8 @@
         "restart status",
         "logs",
         "missing dependency scan"
-      ]
+      ],
+      "healthScope": "readiness"
     }
   ],
   "proves": [
diff --git a/scenarios/plugin-missing-runtime-deps.json b/scenarios/plugin-missing-runtime-deps.json
index 75f8eb5..c75ed92 100644
--- a/scenarios/plugin-missing-runtime-deps.json
+++ b/scenarios/plugin-missing-runtime-deps.json
@@ -31,7 +31,8 @@
         "install result",
         "plugin entry registered",
         "gateway readiness before load"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "restart",
@@ -46,7 +47,8 @@
         "missing dependency diagnostic",
         "plugin load failure",
         "gateway remains supervised"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "survival",
@@ -59,7 +61,8 @@
       "evidence": [
         "status after plugin failure",
         "plugin list after failure"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/plugin-remove.json b/scenarios/plugin-remove.json
index 9bbcd5f..a29b740 100644
--- a/scenarios/plugin-remove.json
+++ b/scenarios/plugin-remove.json
@@ -31,7 +31,8 @@
       "evidence": [
         "install record",
         "plugin appears before uninstall"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "remove",
@@ -46,7 +47,8 @@
         "uninstall output",
         "install index cleanup",
         "registry after removal"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "restart",
@@ -61,7 +63,8 @@
         "restart readiness",
         "removed plugin not loaded",
         "missing dependency scan"
-      ]
+      ],
+      "healthScope": "readiness"
     }
   ],
   "proves": [
diff --git a/scenarios/plugin-update.json b/scenarios/plugin-update.json
index d3784d2..d8cbfd1 100644
--- a/scenarios/plugin-update.json
+++ b/scenarios/plugin-update.json
@@ -31,7 +31,8 @@
       "evidence": [
         "plugin install record",
         "plugin appears in list"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "update",
@@ -46,7 +47,8 @@
         "plugin update dry-run output",
         "tracked plugin metadata",
         "registry refresh"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-update-health",
@@ -60,7 +62,8 @@
         "status after update",
         "plugin lifecycle logs",
         "dependency errors"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/provider-models.json b/scenarios/provider-models.json
index 21bc30f..343c8ec 100644
--- a/scenarios/provider-models.json
+++ b/scenarios/provider-models.json
@@ -28,7 +28,8 @@
         "models list duration",
         "provider timeout warnings",
         "gateway status after model discovery"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "logs",
@@ -41,7 +42,8 @@
         "timeout logs",
         "auth skip logs",
         "gateway stall logs"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/release-runtime-startup.json b/scenarios/release-runtime-startup.json
index 2c61eca..931acaa 100644
--- a/scenarios/release-runtime-startup.json
+++ b/scenarios/release-runtime-startup.json
@@ -36,7 +36,8 @@
         "time to listening",
         "time to health ready",
         "readiness classification"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "post-start",
@@ -52,7 +53,8 @@
         "status command latency",
         "plugin list",
         "plugin startup health"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "startup-logs",
@@ -66,7 +68,8 @@
         "missing dependency errors",
         "plugin service failures",
         "startup phase logs"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/soak.json b/scenarios/soak.json
index 9084fe6..4cdc924 100644
--- a/scenarios/soak.json
+++ b/scenarios/soak.json
@@ -16,7 +16,7 @@
     "soakCommandFailures": 0,
     "soakHealthFailures": 0,
     "rssGrowthMb": 300,
-    "healthP95Ms": 1000
+    "postReadyHealthP95Ms": 1000
   },
   "phases": [
     {
@@ -31,7 +31,8 @@
         "baseline PID",
         "baseline RSS",
         "baseline health"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "loop",
@@ -44,7 +45,8 @@
         "latency trend",
         "RSS trend",
         "logs during loop"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/tui-message-turn.json b/scenarios/tui-message-turn.json
index 7221182..a2dac94 100644
--- a/scenarios/tui-message-turn.json
+++ b/scenarios/tui-message-turn.json
@@ -37,7 +37,8 @@
         "gateway port",
         "runtime binding",
         "env created without service"
-      ]
+      ],
+      "healthScope": "none"
     },
     {
       "id": "gateway-start",
@@ -51,7 +52,8 @@
         "gateway service installed",
         "gateway service started",
         "startup readiness"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "tui-message-turn",
@@ -66,7 +68,8 @@
         "mock provider request timing",
         "gateway health after turn",
         "role resource samples"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-tui-health",
@@ -81,7 +84,8 @@
         "provider logs",
         "plugin errors",
         "memory after TUI turn"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/tui-responsiveness.json b/scenarios/tui-responsiveness.json
index 968f635..bdd93a4 100644
--- a/scenarios/tui-responsiveness.json
+++ b/scenarios/tui-responsiveness.json
@@ -30,7 +30,8 @@
       "evidence": [
         "gateway status",
         "readiness classification"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "tui-smoke",
@@ -43,7 +44,8 @@
         "TUI render time",
         "connected screen",
         "clean interrupt"
-      ]
+      ],
+      "healthScope": "post-ready"
     },
     {
       "id": "post-tui-health",
@@ -57,7 +59,8 @@
         "status after TUI",
         "TUI disconnect logs",
         "gateway errors"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/upgrade-durable-clone-to-local-build.json b/scenarios/upgrade-durable-clone-to-local-build.json
index 04f796b..22e827e 100644
--- a/scenarios/upgrade-durable-clone-to-local-build.json
+++ b/scenarios/upgrade-durable-clone-to-local-build.json
@@ -41,7 +41,8 @@
         "source env",
         "clone root",
         "pre-upgrade service status"
-      ]
+      ],
+      "healthScope": "none"
     },
     {
       "id": "upgrade",
@@ -55,7 +56,8 @@
         "snapshot id",
         "doctor/update output",
         "rollback status"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "post-upgrade",
@@ -74,7 +76,8 @@
         "plugins install index",
         "doctor output",
         "gateway logs without missing dependency/plugin load failures"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/upgrade-existing-user.json b/scenarios/upgrade-existing-user.json
index ac2c990..11ca671 100644
--- a/scenarios/upgrade-existing-user.json
+++ b/scenarios/upgrade-existing-user.json
@@ -28,7 +28,8 @@
         "clone result",
         "source env",
         "clone root"
-      ]
+      ],
+      "healthScope": "none"
     },
     {
       "id": "source-runtime",
@@ -41,7 +42,8 @@
       "evidence": [
         "pre-upgrade runtime",
         "pre-upgrade gateway status"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "upgrade",
@@ -55,7 +57,8 @@
         "snapshot id",
         "doctor/update output",
         "rollback status"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "post-upgrade",
@@ -72,7 +75,8 @@
         "plugins folder/index presence",
         "doctor output",
         "gateway logs"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/upgrade-from-2026-4-20.json b/scenarios/upgrade-from-2026-4-20.json
index 7aa122e..f8add4a 100644
--- a/scenarios/upgrade-from-2026-4-20.json
+++ b/scenarios/upgrade-from-2026-4-20.json
@@ -32,7 +32,8 @@
         "clone result",
         "source env",
         "clone root"
-      ]
+      ],
+      "healthScope": "none"
     },
     {
       "id": "source-runtime",
@@ -47,7 +48,8 @@
         "2026.4.20 upgrade output",
         "pre-upgrade service status",
         "pre-upgrade OpenClaw status"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "upgrade",
@@ -61,7 +63,8 @@
         "snapshot id",
         "doctor/update output",
         "rollback status"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "post-upgrade",
@@ -78,7 +81,8 @@
         "plugins folder/index presence",
         "doctor output",
         "gateway logs"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/upgrade-from-2026-4-24.json b/scenarios/upgrade-from-2026-4-24.json
index f7a5563..50d6d09 100644
--- a/scenarios/upgrade-from-2026-4-24.json
+++ b/scenarios/upgrade-from-2026-4-24.json
@@ -35,7 +35,8 @@
         "clone result",
         "source env",
         "clone root"
-      ]
+      ],
+      "healthScope": "none"
     },
     {
       "id": "source-runtime",
@@ -52,7 +53,8 @@
         "pre-upgrade service status",
         "pre-upgrade OpenClaw status",
         "known 2026.4.24 plugin/runtime-deps logs"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "upgrade",
@@ -66,7 +68,8 @@
         "snapshot id",
         "doctor/update output",
         "rollback status"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "post-upgrade",
@@ -83,7 +86,8 @@
         "plugins install index",
         "doctor output",
         "gateway logs without missing dependency/plugin load failures"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/upgrade-stable-channel-to-beta.json b/scenarios/upgrade-stable-channel-to-beta.json
index ad62a6b..949c68d 100644
--- a/scenarios/upgrade-stable-channel-to-beta.json
+++ b/scenarios/upgrade-stable-channel-to-beta.json
@@ -42,7 +42,8 @@
         "stable channel start output",
         "pre-upgrade gateway status",
         "pre-upgrade OpenClaw status"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "upgrade",
@@ -56,7 +57,8 @@
         "snapshot id",
         "doctor/update output",
         "rollback status"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "post-upgrade",
@@ -75,7 +77,8 @@
         "plugins install index",
         "doctor output",
         "gateway logs without missing dependency/plugin load failures"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/upgrade-stable-channel-to-local-build.json b/scenarios/upgrade-stable-channel-to-local-build.json
index f3669bf..940f47e 100644
--- a/scenarios/upgrade-stable-channel-to-local-build.json
+++ b/scenarios/upgrade-stable-channel-to-local-build.json
@@ -39,7 +39,8 @@
         "stable channel start output",
         "pre-upgrade gateway status",
         "pre-upgrade OpenClaw status"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "upgrade",
@@ -53,7 +54,8 @@
         "snapshot id",
         "doctor/update output",
         "rollback status"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "post-upgrade",
@@ -72,7 +74,8 @@
         "plugins install index",
         "doctor output",
         "gateway logs without missing dependency/plugin load failures"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/scenarios/workspace-scan-pressure.json b/scenarios/workspace-scan-pressure.json
index 99b3ea0..6dddba2 100644
--- a/scenarios/workspace-scan-pressure.json
+++ b/scenarios/workspace-scan-pressure.json
@@ -25,8 +25,8 @@
     "soakCommandFailures": 0,
     "soakHealthFailures": 0,
     "soakHealthP95Ms": 1000,
-    "healthP95Ms": 1000,
-    "peakRssMb": 1000
+    "peakRssMb": 1000,
+    "postReadyHealthP95Ms": 1000
   },
   "phases": [
     {
@@ -41,7 +41,8 @@
         "startup readiness",
         "gateway PID",
         "baseline RSS and CPU"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "restart-after-workspace",
@@ -55,7 +56,8 @@
         "restart readiness",
         "post-fixture gateway status",
         "resource samples during restart"
-      ]
+      ],
+      "healthScope": "readiness"
     },
     {
       "id": "user-facing-commands",
@@ -74,7 +76,8 @@
         "short repeated command p95",
         "health p95",
         "RSS and CPU peaks"
-      ]
+      ],
+      "healthScope": "post-ready"
     }
   ],
   "proves": [
diff --git a/src/evaluator.mjs b/src/evaluator.mjs
index db3a1ac..d1975e0 100644
--- a/src/evaluator.mjs
+++ b/src/evaluator.mjs
@@ -1,6 +1,7 @@
 import { buildAgentTurnBreakdown } from "./collectors/agent-turns.mjs";
 import { computeProviderTurnAttribution } from "./collectors/provider.mjs";
 import { summarizeRuntimeDepsLogs } from "./collectors/logs.mjs";
+import { buildHealthMeasurement, deriveHealthCompatibility } from "./health.mjs";
 import { resolveThresholdPolicy } from "./evaluation/thresholds.mjs";
 import {
   checkAggregateThreshold,
@@ -91,8 +92,15 @@ export function evaluateRecord(record, scenario, options = {}) {
     providerSimulation: agentProviderSimulation
   });
   const finalGatewayState = record.finalMetrics?.service?.gatewayState ?? null;
-  const healthFailures = countHealthFailures(record);
-  const healthP95Ms = collectHealthP95(record);
+  const health = buildHealthMeasurement(record, scenario);
+  const healthCompatibility = deriveHealthCompatibility(health, record);
+  const healthFailures = healthCompatibility.healthFailures;
+  const healthP95Ms = healthCompatibility.healthP95Ms;
+  const startupHealthP95Ms = healthCompatibility.startupHealthP95Ms;
+  const postReadyHealthP95Ms = healthCompatibility.postReadyHealthP95Ms;
+  const startupHealthFailures = healthCompatibility.startupHealthFailures;
+  const postReadyHealthFailures = healthCompatibility.postReadyHealthFailures;
+  const finalHealthFailures = healthCompatibility.finalHealthFailures;
   const soakEvidence = collectSoakEvidence(allResults);
   const mcpBridgeEvidence = collectMcpBridgeEvidence(allResults);
   const browserAutomationEvidence = collectBrowserAutomationEvidence(allResults);
@@ -101,10 +109,10 @@ export function evaluateRecord(record, scenario, options = {}) {
   const officialPluginEvidence = collectOfficialPluginEvidence(allResults);
   const listeningFailures = countListeningFailures(record);
   const tcpConnectMaxMs = collectTcpConnectMax(record);
-  const timeToListeningMs = collectTimeToListening(record);
-  const timeToHealthReadyMs = collectTimeToHealthReady(record);
+  const timeToListeningMs = healthCompatibility.timeToListeningMs ?? collectTimeToListening(record);
+  const timeToHealthReadyMs = healthCompatibility.timeToHealthReadyMs ?? collectTimeToHealthReady(record);
   const readinessFailures = countReadinessFailures(record);
-  const readinessClassification = collectWorstReadinessClassification(record);
+  const readinessClassification = healthCompatibility.readinessClassification ?? collectWorstReadinessClassification(record);
   const coldReadyMs = maxDurationWhere(allResults, (command) => command.startsWith("ocm start "));
   const warmReadyMs = maxDurationWhere(allResults, (command) => command.startsWith("ocm service restart "));
   const upgradeMs = maxDurationWhere(allResults, (command) => command.startsWith("ocm upgrade "));
@@ -203,6 +211,56 @@ export function evaluateRecord(record, scenario, options = {}) {
     });
   }
 
+  if (typeof thresholds.startupHealthFailures === "number" && startupHealthFailures > thresholds.startupHealthFailures) {
+    violations.push({
+      kind: "health",
+      metric: "startupHealthFailures",
+      expected: `<= ${thresholds.startupHealthFailures}`,
+      actual: startupHealthFailures,
+      message: `${startupHealthFailures} startup health check(s) failed, over threshold ${thresholds.startupHealthFailures}`
+    });
+  }
+
+  if (typeof thresholds.postReadyHealthFailures === "number" && postReadyHealthFailures > thresholds.postReadyHealthFailures) {
+    violations.push({
+      kind: "health",
+      metric: "postReadyHealthFailures",
+      expected: `<= ${thresholds.postReadyHealthFailures}`,
+      actual: postReadyHealthFailures,
+      message: `${postReadyHealthFailures} post-ready liveness check(s) failed, over threshold ${thresholds.postReadyHealthFailures}`
+    });
+  }
+
+  if (typeof thresholds.finalHealthFailures === "number" && finalHealthFailures > thresholds.finalHealthFailures) {
+    violations.push({
+      kind: "health",
+      metric: "finalHealthFailures",
+      expected: `<= ${thresholds.finalHealthFailures}`,
+      actual: finalHealthFailures,
+      message: `${finalHealthFailures} final health check(s) failed, over threshold ${thresholds.finalHealthFailures}`
+    });
+  }
+
+  if (typeof thresholds.startupHealthP95Ms === "number" && startupHealthP95Ms !== null && startupHealthP95Ms > thresholds.startupHealthP95Ms) {
+    violations.push({
+      kind: "health",
+      metric: "startupHealthP95Ms",
+      expected: `<= ${thresholds.startupHealthP95Ms}`,
+      actual: startupHealthP95Ms,
+      message: `startup health sample p95 ${startupHealthP95Ms}ms exceeded threshold ${thresholds.startupHealthP95Ms}ms`
+    });
+  }
+
+  if (typeof thresholds.postReadyHealthP95Ms === "number" && postReadyHealthP95Ms !== null && postReadyHealthP95Ms > thresholds.postReadyHealthP95Ms) {
+    violations.push({
+      kind: "health",
+      metric: "postReadyHealthP95Ms",
+      expected: `<= ${thresholds.postReadyHealthP95Ms}`,
+      actual: postReadyHealthP95Ms,
+      message: `post-ready liveness p95 ${postReadyHealthP95Ms}ms exceeded threshold ${thresholds.postReadyHealthP95Ms}ms`
+    });
+  }
+
   if (typeof thresholds.soakMinDurationMs === "number" && soakEvidence.durationMs !== null && soakEvidence.durationMs < thresholds.soakMinDurationMs) {
     violations.push({
       kind: "soak",
@@ -747,6 +805,7 @@ export function evaluateRecord(record, scenario, options = {}) {
     agentProviderRequestCount: providerTurn?.requestCount ?? null,
     agentProviderRequestMissing: providerTurn?.missingProviderRequest ?? null,
     agentProviderAttribution: providerTurn,
+    health,
     tcpConnectMaxMs,
     timeToListeningMs,
     timeToHealthReadyMs,
@@ -758,6 +817,11 @@ export function evaluateRecord(record, scenario, options = {}) {
     finalGatewayState,
     healthFailures,
     healthP95Ms,
+    startupHealthP95Ms,
+    postReadyHealthP95Ms,
+    startupHealthFailures,
+    postReadyHealthFailures,
+    finalHealthFailures,
     soakEvidence,
     mcpBridgeEvidence,
     mcpInitializeMs: mcpBridgeEvidence.initializeMs,
@@ -1882,26 +1946,6 @@ function countGatewayRestarts(record) {
   return commandRestarts + countLogMetric(record, "gatewayRestartMentions");
 }
 
-function collectHealthP95(record) {
-  const p95Values = [];
-  for (const phase of record.phases ?? []) {
-    const p95 = phase.metrics?.healthSummary?.p95Ms;
-    if (typeof p95 === "number") {
-      p95Values.push(p95);
-    }
-  }
-
-  const finalP95 = record.finalMetrics?.healthSummary?.p95Ms;
-  if (typeof finalP95 === "number") {
-    p95Values.push(finalP95);
-  }
-
-  if (p95Values.length === 0) {
-    return null;
-  }
-  return Math.max(...p95Values);
-}
-
 function collectSoakEvidence(results) {
   const loops = results
     .filter((result) => result.command?.includes("run-soak-loop.mjs"))
diff --git a/src/health.mjs b/src/health.mjs
new file mode 100644
index 0000000..fe06b8e
--- /dev/null
+++ b/src/health.mjs
@@ -0,0 +1,319 @@
+export const HEALTH_SCHEMA = "kova.health.v1";
+export const HEALTH_SCOPES = ["readiness", "startup-sample", "post-ready", "final", "none", "unknown"];
+
+const startupScopes = new Set(["readiness", "startup-sample"]);
+
+export function buildHealthMeasurement(record, scenario = null) {
+  const phaseContracts = new Map((scenario?.phases ?? []).map((phase) => [phase.id, phase]));
+  const entries = [];
+  for (const phase of record.phases ?? []) {
+    entries.push({
+      source: "phase",
+      phaseId: phase.id ?? null,
+      scope: normalizeHealthScope(phase.healthScope ?? phaseContracts.get(phase.id)?.healthScope),
+      metrics: phase.metrics ?? null
+    });
+  }
+
+  const finalEntry = {
+    source: "final",
+    phaseId: "final",
+    scope: "final",
+    metrics: record.finalMetrics ?? null
+  };
+  entries.push(finalEntry);
+
+  const readiness = selectReadiness(entries);
+  const startupSamples = summarizeScopedSamples(
+    entries.filter((entry) => startupScopes.has(entry.scope)),
+    "startup-sample",
+    startupSamplesForEntry
+  );
+  const postReadySamples = summarizeScopedSamples(
+    entries.filter((entry) => entry.scope === "post-ready"),
+    "post-ready",
+    postReadySamplesForEntry
+  );
+  const unknownSamples = summarizeScopedSamples(
+    entries.filter((entry) => entry.scope === "unknown"),
+    "unknown",
+    postReadySamplesForEntry
+  );
+  const final = summarizeFinalHealth(finalEntry.metrics);
+  const slowestSample = selectSlowestSample([startupSamples, postReadySamples, final]);
+
+  return {
+    schemaVersion: HEALTH_SCHEMA,
+    readiness,
+    startupSamples,
+    postReadySamples,
+    unknownSamples,
+    final,
+    slowestSample
+  };
+}
+
+export function deriveHealthCompatibility(health, record = null) {
+  const startupHealthP95Ms = health?.startupSamples?.p95Ms ?? null;
+  const postReadyHealthP95Ms = health?.postReadySamples?.p95Ms ?? null;
+  const scopedP95Ms = maxNullable(startupHealthP95Ms, postReadyHealthP95Ms);
+  const oldP95Ms = record ? collectOldHealthP95(record) : null;
+  const startupFailures = health?.startupSamples?.failureCount ?? 0;
+  const postReadyFailures = health?.postReadySamples?.failureCount ?? 0;
+  const unknownFailures = health?.unknownSamples?.failureCount ?? 0;
+  const finalFailures = health?.final?.failureCount ?? 0;
+
+  return {
+    timeToListeningMs: health?.readiness?.listeningReadyAtMs ?? null,
+    timeToHealthReadyMs: health?.readiness?.healthReadyAtMs ?? null,
+    readinessClassification: health?.readiness
+      ? {
+          phaseId: health.readiness.phaseId,
+          state: health.readiness.classification,
+          severity: health.readiness.severity,
+          reason: health.readiness.reason,
+          thresholdMs: health.readiness.thresholdMs,
+          deadlineMs: health.readiness.deadlineMs,
+          listeningReadyAtMs: health.readiness.listeningReadyAtMs,
+          healthReadyAtMs: health.readiness.healthReadyAtMs
+        }
+      : null,
+    healthFailures: startupFailures + postReadyFailures + unknownFailures + finalFailures,
+    healthP95Ms: scopedP95Ms ?? oldP95Ms,
+    startupHealthP95Ms,
+    postReadyHealthP95Ms,
+    startupHealthFailures: startupFailures,
+    postReadyHealthFailures: postReadyFailures,
+    finalHealthFailures: finalFailures
+  };
+}
+
+function normalizeHealthScope(scope) {
+  return typeof scope === "string" && HEALTH_SCOPES.includes(scope) ? scope : "unknown";
+}
+
+function selectReadiness(entries) {
+  const scoped = entries
+    .filter((entry) => startupScopes.has(entry.scope))
+    .map((entry) => readinessValue(entry.metrics?.readiness, entry.phaseId))
+    .filter(Boolean);
+  const candidates = scoped.length > 0
+    ? scoped
+    : entries.map((entry) => readinessValue(entry.metrics?.readiness, entry.phaseId)).filter(Boolean);
+  if (candidates.length === 0) {
+    return null;
+  }
+  candidates.sort((left, right) => {
+    const rankDelta = readinessRank(right.classification) - readinessRank(left.classification);
+    if (rankDelta !== 0) {
+      return rankDelta;
+    }
+    return (right.healthReadyAtMs ?? 0) - (left.healthReadyAtMs ?? 0);
+  });
+  return candidates[0];
+}
+
+function readinessValue(readiness, phaseId) {
+  if (!readiness?.classification || !(readiness.deadlineMs > 0)) {
+    return null;
+  }
+  return {
+    phaseId,
+    listeningReadyAtMs: readiness.listeningReadyAtMs,
+    healthReadyAtMs: readiness.healthReadyAtMs,
+    classification: readiness.classification.state,
+    severity: readiness.classification.severity,
+    reason: readiness.classification.reason,
+    thresholdMs: readiness.thresholdMs,
+    deadlineMs: readiness.deadlineMs,
+    attempts: readiness.attempts ?? null
+  };
+}
+
+function readinessRank(state) {
+  if (state === "hard-failure") {
+    return 4;
+  }
+  if (state === "unhealthy") {
+    return 3;
+  }
+  if (state === "slow-startup") {
+    return 2;
+  }
+  if (state === "ready") {
+    return 1;
+  }
+  return 0;
+}
+
+function startupSamplesForEntry(entry) {
+  const attempts = entry.metrics?.readiness?.healthAttempts;
+  if (Array.isArray(attempts) && attempts.length > 0) {
+    return attempts;
+  }
+  return entry.metrics?.healthSamples ?? [];
+}
+
+function postReadySamplesForEntry(entry) {
+  return entry.metrics?.healthSamples ?? [];
+}
+
+function summarizeScopedSamples(entries, scope, sampleSelector) {
+  const samples = [];
+  for (const entry of entries) {
+    for (const sample of sampleSelector(entry)) {
+      samples.push({ ...sample, phaseId: entry.phaseId });
+    }
+  }
+  if (samples.length > 0) {
+    return summarizeSamples(samples, scope);
+  }
+
+  const summaries = entries
+    .map((entry) => ({ phaseId: entry.phaseId, summary: entry.metrics?.healthSummary }))
+    .filter((entry) => entry.summary);
+  if (summaries.length === 0) {
+    return emptyHealthSummary(scope);
+  }
+
+  let slowestPhaseId = null;
+  let maxMs = null;
+  for (const { phaseId, summary } of summaries) {
+    if (typeof summary.maxMs === "number" && (maxMs === null || summary.maxMs > maxMs)) {
+      maxMs = summary.maxMs;
+      slowestPhaseId = phaseId;
+    }
+  }
+
+  return {
+    scope,
+    count: sum(summaries, "count"),
+    okCount: sum(summaries, "okCount"),
+    failureCount: sum(summaries, "failureCount"),
+    minMs: minNullable(...summaries.map(({ summary }) => summary.minMs)),
+    p50Ms: maxNullable(...summaries.map(({ summary }) => summary.p50Ms)),
+    p95Ms: maxNullable(...summaries.map(({ summary }) => summary.p95Ms)),
+    maxMs,
+    slowestPhaseId
+  };
+}
+
+function summarizeSamples(samples, scope) {
+  const durations = samples
+    .map((sample) => sample.durationMs)
+    .filter((duration) => typeof duration === "number")
+    .sort((left, right) => left - right);
+  let slowestPhaseId = null;
+  let slowestMs = null;
+  for (const sample of samples) {
+    if (typeof sample.durationMs === "number" && (slowestMs === null || sample.durationMs > slowestMs)) {
+      slowestMs = sample.durationMs;
+      slowestPhaseId = sample.phaseId ?? null;
+    }
+  }
+
+  return {
+    scope,
+    count: samples.length,
+    okCount: samples.filter((sample) => sample.ok === true).length,
+    failureCount: samples.filter((sample) => sample.ok !== true).length,
+    minMs: durations.at(0) ?? null,
+    p50Ms: percentile(durations, 0.5),
+    p95Ms: percentile(durations, 0.95),
+    maxMs: durations.at(-1) ?? null,
+    slowestPhaseId
+  };
+}
+
+function emptyHealthSummary(scope) {
+  return {
+    scope,
+    count: 0,
+    okCount: 0,
+    failureCount: 0,
+    minMs: null,
+    p50Ms: null,
+    p95Ms: null,
+    maxMs: null,
+    slowestPhaseId: null
+  };
+}
+
+function summarizeFinalHealth(metrics) {
+  const samples = Array.isArray(metrics?.healthSamples) ? metrics.healthSamples : [];
+  const summary = samples.length > 0 ? summarizeSamples(samples.map((sample) => ({ ...sample, phaseId: "final" })), "final") : null;
+  const fallbackFailureCount = healthFailureCount([metrics?.health]);
+  const failureCount = summary?.failureCount ?? metrics?.healthSummary?.failureCount ?? fallbackFailureCount;
+  const maxMs = summary?.maxMs ?? metrics?.healthSummary?.maxMs ?? metrics?.health?.durationMs ?? null;
+  const p95Ms = summary?.p95Ms ?? metrics?.healthSummary?.p95Ms ?? null;
+  const gatewayState = metrics?.service?.gatewayState ?? null;
+  const ok = metrics
+    ? (gatewayState === null ? failureCount === 0 : gatewayState === "running" && failureCount === 0)
+    : null;
+  return {
+    scope: "final",
+    gatewayState,
+    ok,
+    healthOk: metrics?.health?.ok ?? null,
+    failureCount,
+    p95Ms,
+    maxMs,
+    slowestPhaseId: maxMs === null ? null : "final"
+  };
+}
+
+function selectSlowestSample(summaries) {
+  let slowest = null;
+  for (const summary of summaries) {
+    if (!summary || typeof summary.maxMs !== "number") {
+      continue;
+    }
+    if (!slowest || summary.maxMs > slowest.durationMs) {
+      slowest = {
+        scope: summary.scope,
+        phaseId: summary.slowestPhaseId ?? null,
+        durationMs: summary.maxMs
+      };
+    }
+  }
+  return slowest;
+}
+
+function collectOldHealthP95(record) {
+  const values = [];
+  for (const phase of record?.phases ?? []) {
+    if (typeof phase.metrics?.healthSummary?.p95Ms === "number") {
+      values.push(phase.metrics.healthSummary.p95Ms);
+    }
+  }
+  if (typeof record?.finalMetrics?.healthSummary?.p95Ms === "number") {
+    values.push(record.finalMetrics.healthSummary.p95Ms);
+  }
+  return values.length === 0 ? null : Math.max(...values);
+}
+
+function healthFailureCount(samples) {
+  return samples.filter((sample) => sample && sample.ok === false).length;
+}
+
+function sum(entries, key) {
+  return entries.reduce((total, entry) => total + (entry.summary?.[key] ?? 0), 0);
+}
+
+function maxNullable(...values) {
+  const numeric = values.filter((value) => typeof value === "number");
+  return numeric.length === 0 ? null : Math.max(...numeric);
+}
+
+function minNullable(...values) {
+  const numeric = values.filter((value) => typeof value === "number");
+  return numeric.length === 0 ? null : Math.min(...numeric);
+}
+
+function percentile(values, percentileValue) {
+  if (values.length === 0) {
+    return null;
+  }
+  const index = Math.ceil(values.length * percentileValue) - 1;
+  return values[Math.min(Math.max(index, 0), values.length - 1)];
+}
diff --git a/src/performance/stats.mjs b/src/performance/stats.mjs
index 9a04c92..feac96b 100644
--- a/src/performance/stats.mjs
+++ b/src/performance/stats.mjs
@@ -20,6 +20,8 @@ export const PERFORMANCE_METRICS = [
   { id: "coldPreProviderMs", title: "Cold Pre-Provider", unit: "ms", regressionKey: "agentLatencyRegressionPercent" },
   { id: "warmPreProviderMs", title: "Warm Pre-Provider", unit: "ms", regressionKey: "agentLatencyRegressionPercent" },
   { id: "healthP95Ms", title: "Health p95", unit: "ms", regressionKey: "startupRegressionPercent" },
+  { id: "startupHealthP95Ms", title: "Startup Health p95", unit: "ms", regressionKey: "startupRegressionPercent" },
+  { id: "postReadyHealthP95Ms", title: "Post-Ready Health p95", unit: "ms", regressionKey: "startupRegressionPercent" },
   { id: "runtimeDepsStagingMs", title: "Runtime Deps Staging", unit: "ms", regressionKey: "startupRegressionPercent" }
 ];
 
diff --git a/src/registries/scenarios.mjs b/src/registries/scenarios.mjs
index 48e90ca..c934d29 100644
--- a/src/registries/scenarios.mjs
+++ b/src/registries/scenarios.mjs
@@ -1,6 +1,8 @@
 import { scenariosDir } from "../paths.mjs";
 import { assertNoShapeErrors, loadJsonRegistry, requireArray, requireKebabId, requireObject, requireString } from "./validate.mjs";
 
+export const HEALTH_SCOPES = ["readiness", "startup-sample", "post-ready", "final", "none"];
+
 export async function loadScenarios(selectedId) {
   return loadJsonRegistry({
     dir: scenariosDir,
@@ -108,6 +110,7 @@ function validatePhases(phases, errors) {
     requireKebabId(phase, "id", errors, prefix);
     requireString(phase, "title", errors, prefix);
     requireString(phase, "intent", errors, prefix);
+    requireString(phase, "healthScope", errors, prefix);
     requireArray(phase, "commands", errors, prefix);
     requireArray(phase, "evidence", errors, prefix);
 
@@ -120,6 +123,9 @@ function validatePhases(phases, errors) {
 
     validateStringArray(phase.commands, `${prefix}.commands`, errors);
     validateStringArray(phase.evidence, `${prefix}.evidence`, errors);
+    if (typeof phase.healthScope === "string" && !HEALTH_SCOPES.includes(phase.healthScope)) {
+      errors.push(`${prefix}.healthScope must be one of ${HEALTH_SCOPES.join(", ")}`);
+    }
     if (phase.expectedAgentFailure !== undefined && typeof phase.expectedAgentFailure !== "boolean") {
       errors.push(`${prefix}.expectedAgentFailure must be a boolean when set`);
     }
diff --git a/src/reporting/compare.mjs b/src/reporting/compare.mjs
index 0c07699..6614521 100644
--- a/src/reporting/compare.mjs
+++ b/src/reporting/compare.mjs
@@ -21,6 +21,11 @@ const defaultThresholds = {
   timeToHealthReadyMs: 5000,
   readinessFailures: 0,
   healthP95Ms: 1000,
+  startupHealthFailures: 0,
+  postReadyHealthFailures: 0,
+  finalHealthFailures: 0,
+  startupHealthP95Ms: 1000,
+  postReadyHealthP95Ms: 1000,
   gatewayRestartCount: 0,
   providerTimeoutMentions: 0,
   eventLoopDelayMentions: 0,
@@ -317,6 +322,8 @@ function diagnosticRecordSummary(record) {
     providerFinalMs: measurements.agentProviderFinalMs ?? measurements.coldProviderFinalMs ?? null,
     runtimeDepsStagingMs: measurements.runtimeDepsStagingMs ?? null,
     timeToHealthReadyMs: measurements.timeToHealthReadyMs ?? null,
+    startupHealthP95Ms: measurements.startupHealthP95Ms ?? null,
+    postReadyHealthP95Ms: measurements.postReadyHealthP95Ms ?? null,
     peakRssMb: measurements.peakRssMb ?? null
   };
 }
@@ -406,6 +413,11 @@ function metricDeltas(baseline, current) {
     "timeToHealthReadyMs",
     "healthP95Ms",
     "healthFailures",
+    "startupHealthP95Ms",
+    "postReadyHealthP95Ms",
+    "startupHealthFailures",
+    "postReadyHealthFailures",
+    "finalHealthFailures",
     "readinessFailures",
     "missingDependencyErrors",
     "pluginLoadFailures",
diff --git a/src/reporting/report.mjs b/src/reporting/report.mjs
index 617b733..77ab858 100644
--- a/src/reporting/report.mjs
+++ b/src/reporting/report.mjs
@@ -114,8 +114,7 @@ export function renderMarkdownReport(report) {
       lines.push(`- TCP connect max: ${record.measurements.tcpConnectMaxMs ?? "unknown"} ms`);
       lines.push(`- Missing dependency errors: ${record.measurements.missingDependencyErrors ?? "unknown"}`);
       lines.push(`- Final gateway state: ${record.measurements.finalGatewayState ?? "unknown"}`);
-      lines.push(`- Health failures: ${record.measurements.healthFailures ?? "unknown"}`);
-      lines.push(`- Health p95: ${record.measurements.healthP95Ms ?? "unknown"} ms`);
+      lines.push(...formatHealthMeasurementLines(record.measurements));
       if (record.measurements.soakEvidence?.available) {
         lines.push(`- Soak trend: duration ${record.measurements.soakDurationMs ?? "unknown"} ms; iterations ${record.measurements.soakIterations ?? "unknown"}; command p95 ${record.measurements.soakCommandP95Ms ?? "unknown"} ms; health p95 ${record.measurements.soakHealthP95Ms ?? "unknown"} ms; RSS growth ${record.measurements.rssGrowthMb ?? "unknown"} MB; gateway RSS growth ${record.measurements.gatewayRssGrowthMb ?? "unknown"} MB`);
       }
@@ -499,6 +498,26 @@ function formatMetrics(metrics) {
   return lines.length > 0 ? lines : ["- unavailable"];
 }
 
+function formatHealthMeasurementLines(measurements) {
+  const health = measurements.health;
+  const lines = [
+    `- Health failures: ${measurements.healthFailures ?? "unknown"}`,
+    `- Startup health p95: ${measurements.startupHealthP95Ms ?? health?.startupSamples?.p95Ms ?? "unknown"} ms`,
+    `- Post-ready liveness p95: ${measurements.postReadyHealthP95Ms ?? health?.postReadySamples?.p95Ms ?? "unknown"} ms`,
+    `- Final health failures: ${measurements.finalHealthFailures ?? health?.final?.failureCount ?? "unknown"}`
+  ];
+  if (health?.final) {
+    const healthState = health.final.healthOk === null ? "unknown" : health.final.healthOk ? "ok" : "not-ok";
+    lines.push(`- Final health state: gateway ${health.final.gatewayState ?? "unknown"}; health ${healthState}`);
+  }
+  if (health?.slowestSample) {
+    lines.push(`- Slowest health sample: ${health.slowestSample.scope} ${health.slowestSample.phaseId ?? "unknown"} ${health.slowestSample.durationMs} ms`);
+  } else if (measurements.healthP95Ms !== null && measurements.healthP95Ms !== undefined) {
+    lines.push(`- Compatibility health p95: ${measurements.healthP95Ms} ms`);
+  }
+  return lines;
+}
+
 function formatRecordFailureCards(records = []) {
   const cards = records
     .filter((record) => !["PASS", "DRY-RUN"].includes(record.status))
@@ -670,7 +689,14 @@ function summarizeMeasurements(measurements) {
     timeToHealthReadyMs: measurements.timeToHealthReadyMs ?? null,
     readinessClassification: measurements.readinessClassification ?? null,
     readinessClassificationReason: measurements.readinessClassificationReason ?? null,
+    health: measurements.health ?? null,
     healthFailures: measurements.healthFailures ?? null,
+    healthP95Ms: measurements.healthP95Ms ?? null,
+    startupHealthP95Ms: measurements.startupHealthP95Ms ?? null,
+    postReadyHealthP95Ms: measurements.postReadyHealthP95Ms ?? null,
+    startupHealthFailures: measurements.startupHealthFailures ?? null,
+    postReadyHealthFailures: measurements.postReadyHealthFailures ?? null,
+    finalHealthFailures: measurements.finalHealthFailures ?? null,
     missingDependencyErrors: measurements.missingDependencyErrors ?? null,
     pluginLoadFailures: measurements.pluginLoadFailures ?? null,
     officialPluginEvidence: measurements.officialPluginEvidence ?? null,
@@ -1150,6 +1176,7 @@ function compactRolePeaks(measurements) {
 function pushMeasurementBrief(lines, measurements, { compact }) {
   lines.push("Measurements:");
   lines.push(`- startup: listening ${valueMs(measurements.timeToListeningMs)}; health ${valueMs(measurements.timeToHealthReadyMs)}; readiness ${measurements.readinessClassification ?? "unknown"}; gateway ${measurements.finalGatewayState ?? "unknown"}; restarts ${measurements.gatewayRestartCount ?? "unknown"}`);
+  lines.push(`- health: startup p95 ${valueMs(measurements.startupHealthP95Ms)}; post-ready p95 ${valueMs(measurements.postReadyHealthP95Ms)}; failures ${measurements.healthFailures ?? "unknown"}; final failures ${measurements.finalHealthFailures ?? "unknown"}${healthSlowestText(measurements)}`);
   lines.push(`- resources: peak RSS ${valueMb(measurements.peakRssMb)}; max CPU ${valuePercent(measurements.cpuPercentMax)}; samples ${measurements.resourceSampleCount ?? "unknown"}; roles ${rolePeakText(measurements)}`);
   lines.push(`- agent: turn ${valueMs(measurements.agentTurnMs, "not-run")}; cold/warm ${valueMs(measurements.coldAgentTurnMs)}/${valueMs(measurements.warmAgentTurnMs)}; cold-warm delta ${valueMs(measurements.agentColdWarmDeltaMs)}; pre-provider ${valueMs(measurements.agentPreProviderMs)}; provider ${valueMs(measurements.agentProviderFinalMs)}; cleanup ${valueMs(measurements.agentCleanupMaxMs)}; diagnosis ${measurements.agentLatencyDiagnosis?.kind ?? "unknown"}; leaks ${measurements.agentProcessLeakCount ?? "unknown"}`);
   lines.push(`- plugins/runtime: missing deps ${measurements.missingDependencyErrors ?? "unknown"}; plugin failures ${measurements.pluginLoadFailures ?? "unknown"}; runtime deps ${valueMs(measurements.runtimeDepsStagingMs)}${runtimeDepsPluginText(measurements)}; warm restages ${measurements.warmRuntimeDepsRestageCount ?? "unknown"}; warm reuse ${measurements.runtimeDepsWarmReuseOk ?? "unknown"}`);
@@ -1223,6 +1250,14 @@ function valuePercent(value) {
   return value === null || value === undefined ? "unknown" : `${value}%`;
 }
 
+function healthSlowestText(measurements) {
+  const slowest = measurements.health?.slowestSample;
+  if (!slowest) {
+    return "";
+  }
+  return `; slowest ${slowest.scope}/${slowest.phaseId ?? "unknown"} ${valueMs(slowest.durationMs)}`;
+}
+
 function buildFixerPrompt({ report, primaryBlocker, why, measurements, evidence, likelyOwner }) {
   const parts = [
     `Investigate OpenClaw release gate failure ${primaryBlocker}.`,
diff --git a/src/runner.mjs b/src/runner.mjs
index cd5f017..d6bc1e8 100644
--- a/src/runner.mjs
+++ b/src/runner.mjs
@@ -142,6 +142,7 @@ export async function executeScenario(scenario, context) {
           id: phase.id,
           title: phase.title,
           intent: phase.intent,
+          healthScope: phase.healthScope,
           expectedAgentFailure: phase.expectedAgentFailure === true,
           commands,
           evidence: phase.evidence ?? [],
@@ -347,6 +348,7 @@ function buildPlannedPhases(scenario, context, envName, artifactDir, authPolicy)
       id: phase.id,
       title: phase.title,
       intent: phase.intent,
+      healthScope: phase.healthScope,
       expectedAgentFailure: phase.expectedAgentFailure === true,
       commands: materializeScenarioPhaseCommands(phase, context, envName, artifactDir),
       evidence: phase.evidence ?? []
diff --git a/src/selfcheck.mjs b/src/selfcheck.mjs
index 8c72abb..0b49264 100644
--- a/src/selfcheck.mjs
+++ b/src/selfcheck.mjs
@@ -356,6 +356,8 @@ export async function runSelfCheck(flags = {}) {
     checks.push(markdownFailureCardsCheck());
     checks.push(reportRecommendedNextScenarioCheck());
     checks.push(readinessClassificationCheck());
+    checks.push(healthReadinessModelCheck());
+    checks.push(oldHealthReportCompatibilityCheck());
     checks.push(await resourceRoleAttributionCheck(tmp));
     checks.push(await resourceRootCommandRoleBoundaryCheck());
     checks.push(await resourceRolePollutionCheck());
@@ -365,6 +367,7 @@ export async function runSelfCheck(flags = {}) {
     checks.push(await cleanupRetryCheck(tmp));
     checks.push(stateRegistryValidationCheck());
     checks.push(scenarioCloneFirstValidationCheck());
+    checks.push(scenarioHealthScopeValidationCheck());
     checks.push(scenarioStateCompatibilityCheck());
     checks.push(await cpuProfileParserCheck());
     checks.push(await heapProfileParserCheck());
@@ -3880,6 +3883,7 @@ function readinessClassificationCheck() {
       phases: [
         {
           id: "provision",
+          healthScope: "readiness",
           results: [],
           metrics: {
             readiness: {
@@ -3951,6 +3955,186 @@ function readinessClassificationCheck() {
   }
 }
 
+function healthReadinessModelCheck() {
+  try {
+    const record = {
+      status: "PASS",
+      phases: [
+        {
+          id: "cold-start",
+          healthScope: "readiness",
+          results: [],
+          metrics: {
+            readiness: {
+              deadlineMs: 90000,
+              thresholdMs: 30000,
+              ready: true,
+              listeningReady: true,
+              listeningReadyAtMs: 120,
+              healthReadyAtMs: 200,
+              attempts: 2,
+              classification: {
+                state: "ready",
+                severity: "pass",
+                reason: "gateway became healthy within the readiness threshold"
+              },
+              healthAttempts: [
+                { ok: false, durationMs: 25 },
+                { ok: true, durationMs: 30 }
+              ]
+            },
+            healthSamples: [
+              { ok: true, durationMs: 40 }
+            ],
+            healthSummary: {
+              count: 1,
+              okCount: 1,
+              failureCount: 0,
+              minMs: 40,
+              p50Ms: 40,
+              p95Ms: 40,
+              maxMs: 40
+            }
+          }
+        },
+        {
+          id: "api-latency",
+          healthScope: "post-ready",
+          results: [],
+          metrics: {
+            healthSamples: [
+              { ok: true, durationMs: 10 },
+              { ok: true, durationMs: 1500 }
+            ],
+            healthSummary: {
+              count: 2,
+              okCount: 2,
+              failureCount: 0,
+              minMs: 10,
+              p50Ms: 10,
+              p95Ms: 1500,
+              maxMs: 1500
+            }
+          }
+        }
+      ],
+      finalMetrics: {
+        service: { gatewayState: "running" },
+        healthSamples: [{ ok: true, durationMs: 50 }],
+        healthSummary: {
+          count: 1,
+          okCount: 1,
+          failureCount: 0,
+          minMs: 50,
+          p50Ms: 50,
+          p95Ms: 50,
+          maxMs: 50
+        },
+        health: { ok: true, durationMs: 50 }
+      }
+    };
+    const scenario = {
+      phases: [
+        { id: "cold-start", healthScope: "readiness" },
+        { id: "api-latency", healthScope: "post-ready" }
+      ],
+      thresholds: {
+        gatewayReadyMs: 30000,
+        postReadyHealthP95Ms: 1000
+      }
+    };
+    evaluateRecord(record, scenario);
+    assertEqual(record.status, "FAIL", "post-ready health threshold fails");
+    assertEqual(record.measurements.health.schemaVersion, "kova.health.v1", "health schema");
+    assertEqual(record.measurements.timeToHealthReadyMs, 200, "readiness health ready derived");
+    assertEqual(record.measurements.startupHealthP95Ms, 30, "startup health p95 derived from readiness attempts");
+    assertEqual(record.measurements.postReadyHealthP95Ms, 1500, "post-ready health p95 derived from post-ready samples");
+    assertEqual(record.measurements.healthP95Ms, 1500, "compatibility health p95 derived");
+    assertEqual(record.measurements.health.slowestSample.scope, "post-ready", "slowest health scope");
+    assertEqual(
+      record.violations.some((violation) => violation.metric === "postReadyHealthP95Ms"),
+      true,
+      "post-ready health violation"
+    );
+    assertEqual(
+      record.violations.some((violation) => violation.metric === "timeToHealthReadyMs"),
+      false,
+      "post-ready liveness does not masquerade as readiness"
+    );
+    return {
+      id: "health-readiness-model",
+      status: "PASS",
+      command: "evaluate synthetic scoped health record",
+      durationMs: 0
+    };
+  } catch (error) {
+    return {
+      id: "health-readiness-model",
+      status: "FAIL",
+      command: "evaluate synthetic scoped health record",
+      durationMs: 0,
+      message: error.message
+    };
+  }
+}
+
+function oldHealthReportCompatibilityCheck() {
+  try {
+    const report = {
+      schemaVersion: "kova.report.v1",
+      generatedAt: "2026-05-05T00:00:00.000Z",
+      runId: "old-health-report",
+      mode: "execution",
+      target: "runtime:stable",
+      platform: { os: "darwin", release: "25.0.0", arch: "arm64", node: process.version },
+      summary: { total: 1, statuses: { PASS: 1 } },
+      records: [{
+        scenario: "fresh-install",
+        title: "Fresh Install",
+        status: "PASS",
+        target: "runtime:stable",
+        state: { id: "fresh", title: "Fresh" },
+        envName: "kova-old-health",
+        likelyOwner: "OpenClaw",
+        objective: "Old report compatibility.",
+        measurements: {
+          peakRssMb: 100,
+          cpuPercentMax: 10,
+          timeToListeningMs: 100,
+          timeToHealthReadyMs: 200,
+          readinessClassification: "ready",
+          healthFailures: 0,
+          healthP95Ms: 900,
+          finalGatewayState: "running"
+        },
+        phases: [],
+        violations: []
+      }]
+    };
+    const summary = renderReportSummary(report, { structured: true });
+    assertEqual(summary.scenarios[0].measurements.health, null, "old report health object absent");
+    assertEqual(summary.scenarios[0].measurements.healthP95Ms, 900, "old report health p95 summarized");
+    const markdown = renderMarkdownReport(report);
+    assertEqual(markdown.includes("Compatibility health p95: 900 ms"), true, "old report markdown compatibility p95");
+    const comparison = compareReports(report, report, { thresholds: { healthP95Ms: 0 } });
+    assertEqual(comparison.ok, true, "old report compare remains ok");
+    return {
+      id: "old-health-report-compatibility",
+      status: "PASS",
+      command: "summarize and compare legacy health report shape",
+      durationMs: 0
+    };
+  } catch (error) {
+    return {
+      id: "old-health-report-compatibility",
+      status: "FAIL",
+      command: "summarize and compare legacy health report shape",
+      durationMs: 0,
+      message: error.message
+    };
+  }
+}
+
 async function resourceRoleAttributionCheck(tmp) {
   const command = "node -e 'setTimeout(() => {}, 650)'";
   const artifactPath = join(tmp, "resource-role-attribution.jsonl");
@@ -4739,6 +4923,7 @@ function scenarioCloneFirstValidationCheck() {
           id: "status",
           title: "Status",
           intent: "Unsafe durable source access.",
+          healthScope: "post-ready",
           commands: ["ocm service status {sourceEnv} --json"],
           evidence: ["status"]
         }]
@@ -4762,6 +4947,7 @@ function scenarioCloneFirstValidationCheck() {
           id: "clone",
           title: "Clone",
           intent: "Clone source.",
+          healthScope: "none",
           commands: ["ocm env clone {sourceEnv} {env} --json", "ocm logs {sourceEnv} --tail 20"],
           evidence: ["clone"]
         }]
@@ -4783,12 +4969,14 @@ function scenarioCloneFirstValidationCheck() {
         id: "clone",
         title: "Clone",
         intent: "Clone source.",
+        healthScope: "none",
         commands: ["ocm env clone {sourceEnv} {env} --json"],
         evidence: ["clone"]
       }, {
         id: "upgrade",
         title: "Upgrade",
         intent: "Upgrade disposable clone.",
+        healthScope: "readiness",
         commands: ["ocm upgrade {env} --channel beta --json"],
         evidence: ["upgrade"]
       }]
@@ -4811,6 +4999,72 @@ function scenarioCloneFirstValidationCheck() {
   }
 }
 
+function scenarioHealthScopeValidationCheck() {
+  try {
+    let rejectedMissing = false;
+    try {
+      validateScenarioShape({
+        id: "missing-health-scope",
+        surface: "fresh-install",
+        title: "Missing Health Scope",
+        objective: "Scenario phase without an explicit health scope.",
+        tags: ["fresh-user"],
+        proves: ["baseline"],
+        thresholds: {},
+        phases: [{
+          id: "start",
+          title: "Start",
+          intent: "Start gateway.",
+          commands: ["ocm start {env} {startSelector} --json"],
+          evidence: ["start"]
+        }]
+      }, "missing-health-scope.json");
+    } catch (error) {
+      rejectedMissing = /phases\[0\]\.healthScope must be a non-empty string/.test(error.message);
+    }
+    assertEqual(rejectedMissing, true, "missing healthScope rejected");
+
+    let rejectedInvalid = false;
+    try {
+      validateScenarioShape({
+        id: "invalid-health-scope",
+        surface: "fresh-install",
+        title: "Invalid Health Scope",
+        objective: "Scenario phase with an invalid health scope.",
+        tags: ["fresh-user"],
+        proves: ["baseline"],
+        thresholds: {},
+        phases: [{
+          id: "start",
+          title: "Start",
+          intent: "Start gateway.",
+          healthScope: "startup",
+          commands: ["ocm start {env} {startSelector} --json"],
+          evidence: ["start"]
+        }]
+      }, "invalid-health-scope.json");
+    } catch (error) {
+      rejectedInvalid = /healthScope must be one of/.test(error.message);
+    }
+    assertEqual(rejectedInvalid, true, "invalid healthScope rejected");
+
+    return {
+      id: "scenario-health-scope-validation",
+      status: "PASS",
+      command: "validate scenario health scope contracts",
+      durationMs: 0
+    };
+  } catch (error) {
+    return {
+      id: "scenario-health-scope-validation",
+      status: "FAIL",
+      command: "validate scenario health scope contracts",
+      durationMs: 0,
+      message: error.message
+    };
+  }
+}
+
 function scenarioStateCompatibilityCheck() {
   try {
     let rejected = false;
diff --git a/surfaces/agent-cli-local-turn.json b/surfaces/agent-cli-local-turn.json
index 91b5ef1..030b4f1 100644
--- a/surfaces/agent-cli-local-turn.json
+++ b/surfaces/agent-cli-local-turn.json
@@ -18,8 +18,8 @@
     "preProviderMs": 10000,
     "providerFinalMs": 3000,
     "agentCleanupMs": 5000,
-    "healthP95Ms": 1000,
-    "peakRssMb": 900
+    "peakRssMb": 900,
+    "postReadyHealthP95Ms": 1000
   },
   "roleThresholds": {
     "gateway": {
@@ -81,7 +81,7 @@
         "warmPreProviderMs",
         "agentPreProviderP95Ms",
         "agentCleanupMaxMs",
-        "healthP95Ms",
+        "postReadyHealthP95Ms",
         "peakRssMb",
         "providerTimeoutMentions",
         "pluginLoadFailures"
diff --git a/surfaces/agent-gateway-rpc-turn.json b/surfaces/agent-gateway-rpc-turn.json
index 05439ff..5b87a0a 100644
--- a/surfaces/agent-gateway-rpc-turn.json
+++ b/surfaces/agent-gateway-rpc-turn.json
@@ -14,8 +14,8 @@
     "agentTurnMs": 45000,
     "preProviderMs": 10000,
     "providerFinalMs": 3000,
-    "healthP95Ms": 1000,
-    "peakRssMb": 900
+    "peakRssMb": 900,
+    "postReadyHealthP95Ms": 1000
   },
   "roleThresholds": {
     "gateway": {
@@ -63,7 +63,7 @@
         "agentTurnP95Ms",
         "agentTurnMaxMs",
         "coldPreProviderMs",
-        "healthP95Ms",
+        "postReadyHealthP95Ms",
         "peakRssMb",
         "pluginLoadFailures"
       ]
diff --git a/surfaces/dashboard-session-send-turn.json b/surfaces/dashboard-session-send-turn.json
index 6d0f912..9a73832 100644
--- a/surfaces/dashboard-session-send-turn.json
+++ b/surfaces/dashboard-session-send-turn.json
@@ -14,8 +14,8 @@
     "agentTurnMs": 45000,
     "preProviderMs": 10000,
     "providerFinalMs": 3000,
-    "healthP95Ms": 1000,
-    "peakRssMb": 900
+    "peakRssMb": 900,
+    "postReadyHealthP95Ms": 1000
   },
   "roleThresholds": {
     "gateway": {
@@ -61,7 +61,7 @@
         "agentTurnMs",
         "agentTurnP95Ms",
         "coldPreProviderMs",
-        "healthP95Ms",
+        "postReadyHealthP95Ms",
         "peakRssMb",
         "pluginLoadFailures"
       ]
diff --git a/surfaces/gateway-performance.json b/surfaces/gateway-performance.json
index c4595c3..7a79382 100644
--- a/surfaces/gateway-performance.json
+++ b/surfaces/gateway-performance.json
@@ -12,8 +12,8 @@
   "thresholds": {
     "coldReadyMs": 30000,
     "warmReadyMs": 15000,
-    "healthP95Ms": 1000,
-    "peakRssMb": 900
+    "peakRssMb": 900,
+    "postReadyHealthP95Ms": 1000
   },
   "roleThresholds": {
     "gateway": {
@@ -60,7 +60,7 @@
       "metrics": [
         "coldReadyMs",
         "warmReadyMs",
-        "healthP95Ms",
+        "postReadyHealthP95Ms",
         "peakRssMb",
         "eventLoopMaxMs"
       ]
diff --git a/surfaces/openai-compatible-turn.json b/surfaces/openai-compatible-turn.json
index a16842d..fa18b37 100644
--- a/surfaces/openai-compatible-turn.json
+++ b/surfaces/openai-compatible-turn.json
@@ -14,8 +14,8 @@
     "agentTurnMs": 45000,
     "preProviderMs": 10000,
     "providerFinalMs": 3000,
-    "healthP95Ms": 1000,
-    "peakRssMb": 900
+    "peakRssMb": 900,
+    "postReadyHealthP95Ms": 1000
   },
   "roleThresholds": {
     "gateway": {
@@ -62,7 +62,7 @@
         "agentTurnMs",
         "agentTurnP95Ms",
         "coldPreProviderMs",
-        "healthP95Ms",
+        "postReadyHealthP95Ms",
         "peakRssMb",
         "pluginLoadFailures"
       ]
diff --git a/surfaces/soak.json b/surfaces/soak.json
index cb19e38..b3c947c 100644
--- a/surfaces/soak.json
+++ b/surfaces/soak.json
@@ -19,7 +19,7 @@
     "rssGrowthMb": 300,
     "gatewayRssGrowthMb": 300,
     "soakHealthP95Ms": 1000,
-    "healthP95Ms": 1000
+    "postReadyHealthP95Ms": 1000
   },
   "roleThresholds": {
     "gateway": {
diff --git a/surfaces/tui-message-turn.json b/surfaces/tui-message-turn.json
index 1542e25..59b20df 100644
--- a/surfaces/tui-message-turn.json
+++ b/surfaces/tui-message-turn.json
@@ -14,8 +14,8 @@
     "agentTurnMs": 45000,
     "preProviderMs": 10000,
     "providerFinalMs": 3000,
-    "healthP95Ms": 1000,
-    "peakRssMb": 900
+    "peakRssMb": 900,
+    "postReadyHealthP95Ms": 1000
   },
   "roleThresholds": {
     "gateway": {
@@ -61,7 +61,7 @@
         "agentTurnMs",
         "agentTurnP95Ms",
         "coldPreProviderMs",
-        "healthP95Ms",
+        "postReadyHealthP95Ms",
         "peakRssMb",
         "pluginLoadFailures"
       ]
diff --git a/surfaces/workspace-scan.json b/surfaces/workspace-scan.json
index f51dca9..9201dc4 100644
--- a/surfaces/workspace-scan.json
+++ b/surfaces/workspace-scan.json
@@ -18,9 +18,9 @@
     "modelsListMs": 20000,
     "soakCommandP95Ms": 12000,
     "soakHealthP95Ms": 1000,
-    "healthP95Ms": 1000,
     "peakRssMb": 1000,
-    "eventLoopMaxMs": 500
+    "eventLoopMaxMs": 500,
+    "postReadyHealthP95Ms": 1000
   },
   "roleThresholds": {
     "gateway": {
@@ -74,7 +74,7 @@
         "soakCommandP95Ms",
         "soakHealthP95Ms",
         "peakRssMb",
-        "healthP95Ms",
+        "postReadyHealthP95Ms",
         "eventLoopMaxMs"
       ]
     }