instar 1.3.919 → 1.3.921

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "instar",
-  "version": "1.3.919",
+  "version": "1.3.921",
   "description": "Coherence infrastructure for self-evolving AI agents — on the Claude Code or Codex subscription you already have.",
   "type": "module",
   "main": "dist/index.js",

package/src/data/benchmarkDivergenceRegistry.ts

@@ -58,6 +58,14 @@ export const MODEL_ID_NORMALIZATION: Readonly<Record<string, string>> = {
   'claude-haiku-4-5-20251001': 'claude-haiku-4-5-20251001',
   'gpt-5.5': 'gpt-5.5',
   'gpt-5.5-codex': 'gpt-5.5-codex',
+  // The model actually serving `messaging-tone-gate` in production (verified
+  // 2026-07-23 on the live meter: attribution.models = ['gpt-5.4-mini'] via
+  // codex-cli). Its absence made the highest-volume enrolled pair report
+  // `no-benched-baseline (unmapped)` — i.e. the one pair with real traffic was
+  // structurally unmeasurable. An identity mapping, like the rest: the battery
+  // reports runs under the same immutable public id production records.
+  'gpt-5.4-mini': 'gpt-5.4-mini',
+  'gemini-2.5-flash': 'gemini-2.5-flash',
   'gemini-2.5-pro': 'gemini-2.5-pro',
   'gemini-3.1-pro': 'gemini-3.1-pro',
 };

package/src/data/benchmarkPredictions.json

@@ -0,0 +1,33 @@
+{
+ "_doc": "Content-free INSTAR-Bench predictions mirror (benchmark-divergence-detector FD1). Counts, not just rates, so the prediction carries its own sampling uncertainty. Regenerate by re-running the battery against the CURRENT prompt templates and re-stamping benchedPromptHash.",
+ "capturedAt": "2026-07-23T22:30:00.000Z",
+ "tasks": {
+  "tone-gate": {
+   "perModel": {
+    "gpt-5.4-mini": {
+     "passRate": 1,
+     "passes": 14,
+     "deterministic": 14
+    },
+    "gpt-5.5": {
+     "passRate": 0.9643,
+     "passes": 27,
+     "deterministic": 28
+    },
+    "gemini-2.5-flash": {
+     "passRate": 1,
+     "passes": 14,
+     "deterministic": 14
+    },
+    "claude-haiku-4-5-20251001": {
+     "passRate": 0.9286,
+     "passes": 13,
+     "deterministic": 14
+    }
+   },
+   "benchedPromptSource": "src/core/MessagingToneGate.ts:TONE_GATE_PROMPT_TEMPLATE",
+   "benchedPromptHash": "4ae0cf6b02663c21a695393727bd05662334f0d8f1a9d3e45d8d9b2141208d54",
+   "capturedAt": "2026-07-23T22:30:00.000Z"
+  }
+ }
+}

package/src/data/builtin-manifest.json

@@ -1,8 +1,8 @@
 {
   "$schema": "./builtin-manifest.schema.json",
   "schemaVersion": 1,
-  "generatedAt": "2026-07-23T21:53:49.579Z",
-  "instarVersion": "1.3.919",
+  "generatedAt": "2026-07-23T23:04:10.011Z",
+  "instarVersion": "1.3.921",
   "entryCount": 202,
   "entries": {
     "hook:session-start": {
@@ -418,7 +418,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:agents": {
@@ -426,7 +426,7 @@
       "type": "route-group",
       "domain": "sessions",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:backups": {
@@ -434,7 +434,7 @@
       "type": "route-group",
       "domain": "operations",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:git": {
@@ -442,7 +442,7 @@
       "type": "route-group",
       "domain": "coordination",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:memory": {
@@ -450,7 +450,7 @@
       "type": "route-group",
       "domain": "memory",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:semantic": {
@@ -458,7 +458,7 @@
       "type": "route-group",
       "domain": "memory",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:status": {
@@ -466,7 +466,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:capabilities": {
@@ -474,7 +474,7 @@
       "type": "route-group",
       "domain": "mapping",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:project-map": {
@@ -482,7 +482,7 @@
       "type": "route-group",
       "domain": "mapping",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:coherence": {
@@ -490,7 +490,7 @@
       "type": "route-group",
       "domain": "coherence",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:topic-bindings": {
@@ -498,7 +498,7 @@
       "type": "route-group",
       "domain": "sessions",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:context": {
@@ -506,7 +506,7 @@
       "type": "route-group",
       "domain": "context",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:scope-coherence": {
@@ -514,7 +514,7 @@
       "type": "route-group",
       "domain": "coherence",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:canonical-state": {
@@ -522,7 +522,7 @@
       "type": "route-group",
       "domain": "state",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:ci": {
@@ -530,7 +530,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:sessions": {
@@ -538,7 +538,7 @@
       "type": "route-group",
       "domain": "sessions",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:jobs": {
@@ -546,7 +546,7 @@
       "type": "route-group",
       "domain": "scheduling",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:skip-ledger": {
@@ -554,7 +554,7 @@
       "type": "route-group",
       "domain": "scheduling",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:telegram": {
@@ -562,7 +562,7 @@
       "type": "route-group",
       "domain": "communication",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:attention": {
@@ -570,7 +570,7 @@
       "type": "route-group",
       "domain": "communication",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:relationships": {
@@ -578,7 +578,7 @@
       "type": "route-group",
       "domain": "relationships",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:feedback": {
@@ -586,7 +586,7 @@
       "type": "route-group",
       "domain": "feedback",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:updates": {
@@ -594,7 +594,7 @@
       "type": "route-group",
       "domain": "updates",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:dispatches": {
@@ -602,7 +602,7 @@
       "type": "route-group",
       "domain": "dispatches",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:quota": {
@@ -610,7 +610,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:publishing": {
@@ -618,7 +618,7 @@
       "type": "route-group",
       "domain": "publishing",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:private-views": {
@@ -626,7 +626,7 @@
       "type": "route-group",
       "domain": "publishing",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:tunnel": {
@@ -634,7 +634,7 @@
       "type": "route-group",
       "domain": "networking",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:events": {
@@ -642,7 +642,7 @@
       "type": "route-group",
       "domain": "networking",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:evolution": {
@@ -650,7 +650,7 @@
       "type": "route-group",
       "domain": "evolution",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:watchdog": {
@@ -658,7 +658,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:topic-memory": {
@@ -666,7 +666,7 @@
       "type": "route-group",
       "domain": "memory",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:state-sync": {
@@ -674,7 +674,7 @@
       "type": "route-group",
       "domain": "coordination",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:intent": {
@@ -682,7 +682,7 @@
       "type": "route-group",
       "domain": "intent",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:triage": {
@@ -690,7 +690,7 @@
       "type": "route-group",
       "domain": "safety",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:operations": {
@@ -698,7 +698,7 @@
       "type": "route-group",
       "domain": "safety",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:sentinel": {
@@ -706,7 +706,7 @@
       "type": "route-group",
       "domain": "safety",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:trust": {
@@ -714,7 +714,7 @@
       "type": "route-group",
       "domain": "safety",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:monitoring": {
@@ -722,7 +722,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:commitments": {
@@ -730,7 +730,7 @@
       "type": "route-group",
       "domain": "commitments",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:episodes": {
@@ -738,7 +738,7 @@
       "type": "route-group",
       "domain": "memory",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:messages": {
@@ -746,7 +746,7 @@
       "type": "route-group",
       "domain": "coordination",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:system-reviews": {
@@ -754,7 +754,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "route-group:machine-mesh": {
@@ -770,7 +770,7 @@
       "type": "route-group",
       "domain": "security",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "0ce8b400cb92f71370dc90e4cce3f5ce2c301b6bd07718a7f54cfebc01d675e0",
+      "contentHash": "d513840a029b00aff6c0a1143122d04d18a3b0d57c58c1d98827c8e54f6abe61",
       "since": "2025-01-01"
     },
     "cli:init": {

package/upgrades/1.3.921.md

@@ -0,0 +1,85 @@
+# Upgrade Guide — vNEXT
+
+<!-- assembled-by: assemble-next-md -->
+<!-- bump: patch -->
+
+## What Changed
+
+The Benchmark-Divergence Detector now ships with its first captured predictions baseline (`src/data/benchmarkPredictions.json`), so `GET /benchmark-divergence` reports `mirror.present: true` instead of degrading every finding to `precondition-failed`.
+
+Two model ids were added to the fail-closed normalization table, including `gpt-5.4-mini` — the model actually serving `messaging-tone-gate` in production. Its absence meant the one enrolled pair with real traffic resolved as unmapped indefinitely, so the comparator was quietly measuring nothing there.
+
+The baseline was produced by re-running the battery against the CURRENT prompt templates rather than reusing July's numbers: the tone-gate prompt had grown from ~35,500 to ~39,900 characters since that run, so the old results would have been stale-on-arrival and correctly suppressed by the template-hash precondition.
+
+Unfinished sign-ins regain a live backing process after server restart. Codes
+submitted to a dead or vanished flow now receive an honest expired response,
+with a fresh replacement created when the account still needs sign-in.
+
+## What to Tell Your User
+
+Nothing to do. The model-quality benchmark now has a current baseline captured against the instructions actually running today, rather than a version from three weeks ago. One honest caveat: this is half of a two-part loop — the comparison also needs real-world outcome grades, which don't exist for this scenario yet, so the drift check stays quiet until that lands.
+
+If a restart interrupts sign-in, the dashboard restores a fresh flow instead of
+showing a technical submission failure.
+
+## Summary of New Capabilities
+
+- `GET /benchmark-divergence` has a real baseline to compare against, with per-model counts (not just rates) so prediction-side sampling uncertainty is weighable.
+
+- Restores unfinished login flows during server boot.
+- Replaces dead flows without ever typing their stale code.
+- Shows explicit expired and fresh-ready states in the dashboard.
+
+## Evidence
+
+**Before** — `GET /benchmark-divergence` on the serving machine (2026-07-23, v1.3.914):
+
+```
+{"mirror":{"present":false,"capturedAt":null,"staleDays":null,"stale":true},
+ "summary":{"byVerdict":{},"unmappedModels":[],"missingModelShare":0.02},"findings":[]}
+```
+
+Every comparison degraded to `precondition-failed`; nothing could be measured.
+
+**The unmapped-model defect** — the live meter reports the enrolled pair's
+production model:
+
+```
+messaging-tone-gate → attribution.models: ["gpt-5.4-mini"], frameworks: ["codex-cli"]
+```
+
+`gpt-5.4-mini` was absent from `MODEL_ID_NORMALIZATION`, and FD5 is fail-closed
+exact-match, so that pair resolved `no-benched-baseline (unmapped)` indefinitely.
+
+**The staleness that made a back-fill useless** — benched vs live template size for
+the two enrolled tasks:
+
+| task | benched | live | match? |
+|---|---|---|---|
+| tone-gate | 35,499 chars | 39,886 chars | no |
+| zombie-classify | 2,029 chars | 1,373 chars | no |
+
+Building a mirror from July's numbers would have carried a benched-prompt hash that
+never matches the live template, so the Q0 precondition would have suppressed every
+finding — a captured-looking baseline that measures nothing.
+
+**After** — the battery re-run against the regenerated template (14 deterministic
+cases per route, subscription doors), loaded back through the real
+`loadBenchmarkMirror`:
+
+```
+present=true capturedAt=2026-07-23T22:30:00.000Z
+tasks=tone-gate
+perModel=gpt-5.4-mini,gpt-5.5,gemini-2.5-flash,claude-haiku-4-5-20251001
+benchedPromptHash=4ae0cf6b02663c21…  (= sha256 of the LIVE template)
+normalize(gpt-5.4-mini)=gpt-5.4-mini
+```
+
+Pass rates against the current prompt: `gpt-5.4-mini` 14/14, `gemini-2.5-flash`
+14/14, `gpt-5.5` 27/28, `claude-haiku-4-5` 13/14. Notably `gpt-5.4-mini` scored
+6/14 against the OLD template in July — the prompt improved substantially, which is
+the opposite of the decay the staleness concern assumed.
+
+**Honest limit:** the drift comparison still cannot fire, because the real-world
+side of it (`messaging-tone-gate` outcome grades) is all-`unknown`. This change
+makes the prediction side current and measurable; it does not close the loop.

package/upgrades/side-effects/benchmark-baseline-first-capture.md

@@ -0,0 +1,89 @@
+# Side-effects review — first benchmark predictions baseline
+
+**Change:** ships `src/data/benchmarkPredictions.json` (the first captured
+INSTAR-Bench predictions mirror) and adds two rows to `MODEL_ID_NORMALIZATION`.
+
+## What actually changed
+
+1. `src/data/benchmarkPredictions.json` — new, content-free: per (task, model)
+   `passRate`/`passes`/`deterministic`, the benched prompt source, and the benched
+   prompt hash. No prompts, no case text, no outputs.
+2. `MODEL_ID_NORMALIZATION` — adds `gpt-5.4-mini` and `gemini-2.5-flash`, both
+   identity mappings.
+3. `tests/unit/benchmarkDivergenceRegistry.test.ts` — the pin that asserted the
+   mirror is ABSENT now asserts it is PRESENT and well-formed.
+
+## Why (two real defects, not just "capture a baseline")
+
+**(a) The comparator had no baseline at all.** `GET /benchmark-divergence`
+reported `mirror.present: false`, so every finding degraded to
+`precondition-failed`. Measured on the serving machine 2026-07-23.
+
+**(b) The production model was unmapped — the one pair with traffic was
+structurally unmeasurable.** The live meter reports `messaging-tone-gate` running
+on `gpt-5.4-mini` via codex-cli. That id was absent from the normalization table,
+and FD5 is (correctly) fail-closed exact-match, so the pair resolved
+`no-benched-baseline (unmapped)` — indefinitely, and quietly. A comparator that
+reports "nothing to compare" forever looks identical to one reporting health.
+
+**(c) The pre-existing bench results were stale against a drifted prompt.** The
+July battery scored a tone-gate prompt of 35,499 chars; the live template is
+39,886. Building a mirror from those numbers would have been stale-on-arrival —
+the Q0 template-hash precondition would have (correctly) suppressed every finding.
+So the battery was **re-run against the current template**, not back-filled.
+
+## How the numbers were produced
+
+The bench task's template file was regenerated from the live
+`TONE_GATE_PROMPT_TEMPLATE` export: verbatim rules body + the harness's single
+per-case slot, which is exactly the declared `promptFidelity` for this task. The
+battery then ran 14 deterministic cases per route.
+
+`benchedPromptHash` is the sha256 of the LIVE exported template, so the detector's
+Q0 precondition passes only while the prompt genuinely matches — and starts failing
+the moment it drifts again, which is the intended tripwire.
+
+## Blast radius
+
+- **Surface:** one read-only observability route (`GET /benchmark-divergence`).
+- **Authority:** none. Findings are `advisory: true` by construction; nothing gates.
+- **Data:** a static JSON file read through existing FD9 clamps. A malformed mirror
+  is already a first-class `present: false` state, never a throw.
+- **Fail direction unchanged:** an unmapped model still resolves fail-closed.
+
+## Risk analysis
+
+**Could this manufacture a false finding?** Only if the real-world side had a rate
+to compare — it does not. `messaging-tone-gate` currently settles every decision
+`unknown`, so the detector has no real grade-rate and every comparison stays
+non-actionable. **This ships one half of a two-halves loop, and that is stated
+rather than implied**: the mirror makes the benchmark current; it does not make the
+drift comparison useful until outcome grading lands.
+
+**Sample size honesty.** 14 deterministic cases per model is a small battery. FD3's
+Wilson half-width term already accounts for that on the prediction side (a 14-case
+battery carries roughly ±0.25 at p=0.5), so a small battery *widens* the flag
+threshold rather than manufacturing divergence. The counts — not just the rates —
+are recorded precisely so that math can run.
+
+**Cost.** The run was intended to use subscription doors only. One metered
+OpenRouter route matched the inclusion filter by substring (`claude-haiku` matched
+`or-claude-haiku-45`) and billed **$0.87** against a key with a $25 daily / $60
+lifetime cap. Disclosed rather than absorbed; that route's result is EXCLUDED from
+the mirror, which carries subscription-door results only.
+
+## Migration parity
+
+None required. A shipped data file and a source constant; existing agents pick both
+up with the normal version bump. No installed-file, config, hook, or template
+surface.
+
+## Testing
+
+Registry suite (14), core (unchanged), analyzer, routes, E2E alive, migration and
+job-template suites — 97 green. `tsc --noEmit` clean.
+
+## Rollback
+
+Delete the JSON (the loader returns to `present: false`, today's behaviour) or
+revert the commit. Nothing persists; no state to unwind.

package/upgrades/side-effects/login-flow-restart-recovery.md

@@ -0,0 +1,38 @@
+# Login Flow Restart Recovery — Side Effects
+
+## Runtime
+
+- Startup re-drives each durable, non-terminal pending login before serving
+  dashboard traffic.
+- A dead-pane submit may create a fresh provider login process and public
+  verification artifact.
+- Missing flows auto-mint only when the matching pool account is explicitly
+  `needs-reauth` or in an owner-login-required drift episode.
+
+## External
+
+Provider login commands may run once per unfinished flow after server restart.
+Background restoration never opens a browser. No stale submitted code is typed,
+stored, returned, or logged.
+
+## 6b. Operator-surface quality
+
+The affected dashboard cell continues to lead with its primary sign-in action.
+The new states use plain language (“expired” and “a fresh sign-in is ready”),
+with no status codes, process names, paths, or internal identifiers as primary
+content. No destructive action is added or promoted. The copy is short enough
+to read at phone width and directs the operator to the fresh flow already shown
+in the same account cell.
+
+## Rollback
+
+Reverting the wizard, boot wiring, route, and dashboard changes restores the
+prior behavior. The pending-login file schema is unchanged.
+
+## Class-Closure Declaration
+
+- Defect class: `unbounded-self-action`
+- Closure: `n/a`
+- Reason: boot recovery is a one-shot pass over the finite durable pending-login
+  set, with at most one re-drive per incomplete record and no self-rescheduling
+  path. Dead-flow refresh is user-triggered and bounded to one flow.