instar 1.3.25 → 1.3.27

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "instar",
-  "version": "1.3.25",
+  "version": "1.3.27",
   "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/builtin-manifest.json

@@ -1,8 +1,8 @@
 {
   "$schema": "./builtin-manifest.schema.json",
   "schemaVersion": 1,
-  "generatedAt": "2026-05-27T15:55:54.557Z",
-  "instarVersion": "1.3.25",
+  "generatedAt": "2026-05-27T17:04:17.011Z",
+  "instarVersion": "1.3.27",
   "entryCount": 192,
   "entries": {
     "hook:session-start": {
@@ -401,7 +401,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:agents": {
@@ -409,7 +409,7 @@
       "type": "route-group",
       "domain": "sessions",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:backups": {
@@ -417,7 +417,7 @@
       "type": "route-group",
       "domain": "operations",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:git": {
@@ -425,7 +425,7 @@
       "type": "route-group",
       "domain": "coordination",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:memory": {
@@ -433,7 +433,7 @@
       "type": "route-group",
       "domain": "memory",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:semantic": {
@@ -441,7 +441,7 @@
       "type": "route-group",
       "domain": "memory",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:status": {
@@ -449,7 +449,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:capabilities": {
@@ -457,7 +457,7 @@
       "type": "route-group",
       "domain": "mapping",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:project-map": {
@@ -465,7 +465,7 @@
       "type": "route-group",
       "domain": "mapping",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:coherence": {
@@ -473,7 +473,7 @@
       "type": "route-group",
       "domain": "coherence",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:topic-bindings": {
@@ -481,7 +481,7 @@
       "type": "route-group",
       "domain": "sessions",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:context": {
@@ -489,7 +489,7 @@
       "type": "route-group",
       "domain": "context",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:scope-coherence": {
@@ -497,7 +497,7 @@
       "type": "route-group",
       "domain": "coherence",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:canonical-state": {
@@ -505,7 +505,7 @@
       "type": "route-group",
       "domain": "state",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:ci": {
@@ -513,7 +513,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:sessions": {
@@ -521,7 +521,7 @@
       "type": "route-group",
       "domain": "sessions",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:jobs": {
@@ -529,7 +529,7 @@
       "type": "route-group",
       "domain": "scheduling",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:skip-ledger": {
@@ -537,7 +537,7 @@
       "type": "route-group",
       "domain": "scheduling",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:telegram": {
@@ -545,7 +545,7 @@
       "type": "route-group",
       "domain": "communication",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:attention": {
@@ -553,7 +553,7 @@
       "type": "route-group",
       "domain": "communication",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:relationships": {
@@ -561,7 +561,7 @@
       "type": "route-group",
       "domain": "relationships",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:feedback": {
@@ -569,7 +569,7 @@
       "type": "route-group",
       "domain": "feedback",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:updates": {
@@ -577,7 +577,7 @@
       "type": "route-group",
       "domain": "updates",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:dispatches": {
@@ -585,7 +585,7 @@
       "type": "route-group",
       "domain": "dispatches",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:quota": {
@@ -593,7 +593,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:publishing": {
@@ -601,7 +601,7 @@
       "type": "route-group",
       "domain": "publishing",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:private-views": {
@@ -609,7 +609,7 @@
       "type": "route-group",
       "domain": "publishing",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:tunnel": {
@@ -617,7 +617,7 @@
       "type": "route-group",
       "domain": "networking",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:events": {
@@ -625,7 +625,7 @@
       "type": "route-group",
       "domain": "networking",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:evolution": {
@@ -633,7 +633,7 @@
       "type": "route-group",
       "domain": "evolution",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:watchdog": {
@@ -641,7 +641,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:topic-memory": {
@@ -649,7 +649,7 @@
       "type": "route-group",
       "domain": "memory",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:state-sync": {
@@ -657,7 +657,7 @@
       "type": "route-group",
       "domain": "coordination",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:intent": {
@@ -665,7 +665,7 @@
       "type": "route-group",
       "domain": "intent",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:triage": {
@@ -673,7 +673,7 @@
       "type": "route-group",
       "domain": "safety",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:operations": {
@@ -681,7 +681,7 @@
       "type": "route-group",
       "domain": "safety",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:sentinel": {
@@ -689,7 +689,7 @@
       "type": "route-group",
       "domain": "safety",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:trust": {
@@ -697,7 +697,7 @@
       "type": "route-group",
       "domain": "safety",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:monitoring": {
@@ -705,7 +705,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:commitments": {
@@ -713,7 +713,7 @@
       "type": "route-group",
       "domain": "commitments",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:episodes": {
@@ -721,7 +721,7 @@
       "type": "route-group",
       "domain": "memory",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:messages": {
@@ -729,7 +729,7 @@
       "type": "route-group",
       "domain": "coordination",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:system-reviews": {
@@ -737,7 +737,7 @@
       "type": "route-group",
       "domain": "monitoring",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "route-group:machine-mesh": {
@@ -753,7 +753,7 @@
       "type": "route-group",
       "domain": "security",
       "sourcePath": "src/server/routes.ts",
-      "contentHash": "c81e7535459969d930c59d73269316d8f6b2f84c1dacb2e2ea2f83969ec768b5",
+      "contentHash": "8acca5f08e7a2fab53ca75a730c1a88e41f9c43fbaae8e6a96417a7e7124acc2",
       "since": "2025-01-01"
     },
     "cli:init": {
@@ -1449,7 +1449,7 @@
       "type": "subsystem",
       "domain": "server",
       "sourcePath": "src/server/AgentServer.ts",
-      "contentHash": "d6d5166ae7e59805087dda14472dae74aa96827082c1d5fd4f4c3bc84b358965",
+      "contentHash": "1ea8e7160275776b2e410635588d2db3cc56083324dff60488cc5e3a41cf2025",
       "since": "2025-01-01"
     },
     "subsystem:session-manager": {

package/src/scaffold/templates/jobs/instar/failure-analyzer.md

@@ -0,0 +1,32 @@
+---
+name: Failure-Learning Analyzer
+description: "Weekly scan of the failure ledger for dev-process patterns. Discovers support-and-diversity-thresholded insights, opens human-approved tracked improvements (never auto-implements), and runs the verify step on past fixes. Off by default; turns on with monitoring.failureLearning.enabled. Tier-1 supervised (this haiku job wraps the deterministic /failures/analyze endpoint and validates each surfaced insight against its evidence before posting). Spec docs/specs/FAILURE-LEARNING-LOOP-SPEC.md sections 4.4 and 4.6.1."
+schedule: "0 9 * * 3"
+priority: medium
+expectedDurationMinutes: 2
+model: haiku
+enabled: false
+tags:
+  - cat:learning
+  - failure-learning
+  - audit
+toolAllowlist: "*"
+unrestrictedTools: true
+---
+Run the weekly Failure-Learning analysis. This job exists because individual failures get fixed and forgotten — this turns the accumulated, attributed record into process-level insight, and closes the loop all the way to a human-approved fix and a verification that the fix worked.
+
+The job:
+
+1. Confirm the feature is on. If `monitoring.failureLearning.enabled` is not true, exit silently — there is nothing to do.
+
+2. Call the analyzer + closed-loop tick:
+   `curl -s -X POST -H "Authorization: Bearer $AUTH" http://localhost:$PORT/failures/analyze`
+   The response is `{ analysis: { insightsDiscovered, clustersConsidered, clustersBelowThreshold }, actedOn, verified }`. The endpoint does the deterministic work: it only surfaces a pattern that crosses the support + source-diversity gate (so a single noisy session or a flaky test can never manufacture one), opens a tracked Evolution Action + a draft Initiative for each genuinely-new insight (it can NEVER change the process itself — a human approves), and runs the verify step on past fixes whose window has elapsed.
+
+3. **Tier-1 supervision (your job).** For each insight in `analysis.insightsDiscovered`, sanity-check it against its own evidence before you surface it: does the recommendation actually match the category and the supporting-failure count? If an insight looks malformed or unsupported, do NOT surface it — note it and move on. You are the validation wrapper around the deterministic core.
+
+4. Surface ONLY genuinely-new, decision-bearing insights, and ONLY when `monitoring.failureLearning.insightTelegramEscalation` is true. Post ONE consolidated, plain-English message to the existing system topic — never a new per-feature topic, never a ping per failure. Lead with the pattern and the recommended process change: "We've now seen 5 concurrency failures across 4 different changes — they cluster in work that skipped a concurrency review. Recommend adding a concurrency checklist to the adversarial pass. I've opened a draft for your approval." If `insightTelegramEscalation` is false (the default), stay silent — the Process Health view is the pull surface; do not push.
+
+5. If `verified` > 0, the verify step concluded on one or more past fixes. These update the Process Health board silently (effective / ineffective / inconclusive). Mention a verified-effective result only if you're already surfacing other insights this run — "and the convergence-checklist change from last month looks like it worked: that failure class dropped." Never claim causation — it's correlational.
+
+IMPORTANT: This is a guardian job, not a doer. It surfaces patterns and opens human-approved drafts. It NEVER changes a skill, a spec, or the process on its own — that's a deliberate human decision after reading the evidence. Plain English, under 600 characters, write like you're texting a teammate about a pattern you noticed in the data.

package/src/scaffold/templates.ts

@@ -462,6 +462,12 @@ This routes feedback to the Instar maintainers automatically. Valid types: \`bug
 - The PromiseBeacon fires cadenced heartbeats on open commitments so you actually follow through (and surfaces atRisk items), and the commitment-check job surfaces overdue ones.
 - **When to use** (PROACTIVE — this is the trigger): the moment you promise the user a future action, open a commitment. NEVER improvise the follow-through with a raw \`sleep\`/background timer or by "remembering" — those do not survive a session ending, a restart, or compaction, so the promise is silently dropped. A registered commitment is the ONLY durable path. (This is distinct from the Evolution Action Queue / \`/commit-action\`, which tracks self-improvement items, not promises to the user.)
 
+**Failure-Learning Loop** — Dev-process failure forensics (instar self-hosting). When something you built breaks later, it's captured and traced back to the spec/initiative/project AND the dev toolchain that produced it; the analyzer surfaces process-gap patterns and opens human-approved tracked fixes, then verifies whether each fix actually reduced that failure class. Ships OFF (\`monitoring.failureLearning.enabled\`); registers itself on the initiative board.
+- Why features break / failure rate by tool / are our fixes working: \`curl -H "Authorization: Bearer $AUTH" http://localhost:${port}/failures/analysis\` — answer from here, never from memory.
+- List / inspect failures: \`curl -H "Authorization: Bearer $AUTH" http://localhost:${port}/failures\` · \`GET /failures/:id\` · discovered insights: \`GET /failures/insights\`
+- File a diagnosis (one-tap): \`curl -X POST -H "Authorization: Bearer $AUTH" -H "X-Instar-Request: 1" http://localhost:${port}/failures -H 'Content-Type: application/json' -d '{"summary":"<what broke>","initiativeId":"<board id it traces to>","severity":"medium"}'\`
+- **When to use** (PROACTIVE): when you diagnose a bug that traces to past work, file it so the pattern can be learned. The loop NEVER changes the process on its own — it opens a draft for your approval. It can never auto-implement (it never creates the record type the autonomous approver acts on). Distinct from the Evolution Action Queue (this produces the evidence + diagnosis that justify an action).
+
 **Cloudflare Tunnel** — Expose the local server to the internet via Cloudflare. Enables remote access to private views, the API, and file serving.
 - Status: \`curl -H "Authorization: Bearer $AUTH" http://localhost:${port}/tunnel\`
 - Configure in \`.instar/config.json\`: \`{"tunnel": {"enabled": true, "type": "quick"}}\`
@@ -678,6 +684,7 @@ I maintain registries that are the source of truth for specific categories. Thes
 |----------|-------------|
 | What can I do? | \`curl -H "Authorization: Bearer $AUTH" http://localhost:${port}/capabilities\` |
 | What are we working on? / status of a project or initiative? | \`curl -H "Authorization: Bearer $AUTH" http://localhost:${port}/initiatives\` + \`/projects\` (and \`/initiatives/digest\` for what needs a decision) — NEVER answer this from memory |
+| Why do features keep breaking? / our failure rate by build skill? / are our process fixes working? | \`curl -H "Authorization: Bearer $AUTH" http://localhost:${port}/failures/analysis\` + \`/failures\` (Failure-Learning Loop — instar dev-process forensics) — NEVER answer this from memory |
 | Who do I work with? | \`.instar/USER.md\` |
 | What have I learned? | \`.instar/MEMORY.md\` |
 | What jobs do I have? | \`.instar/jobs.json\` or \`curl -H "Authorization: Bearer $AUTH" http://localhost:${port}/jobs\` |

package/upgrades/1.3.26.md

@@ -0,0 +1,25 @@
+# Upgrade Guide — vNEXT
+
+<!-- bump: patch -->
+
+## What Changed
+
+Twelfth increment of the **Feedback Factory Migration** (Dawn → Echo; spec `docs/specs/feedback-factory-migration.md`, approved). Ports the **dispatch request handlers** — the "send guidance back to agents" side — out of the reference Next.js endpoint (`the-portal/pages/api/instar/dispatches/index.ts`) into framework-agnostic functions at `src/feedback-factory/dispatch/handlers.ts`, plus the store operations they need (`listDispatches` / `findDispatchByTitle` / `createDispatch`).
+
+List gates on the agent fingerprint, applies the since/type filter and the version-compat filter (so guidance only reaches the agent versions it applies to), and returns the mapped result. Create requires the internal key, validates the body, dedups by title, and creates. **Not wired into any route yet** — no behavioral change. This completes the front door's request layer (receive a report + send guidance both ported).
+
+## What to Tell Your User
+
+- Both halves of the feedback loop's front door are now ported and behave exactly like Dawn's originals — catching reports, and sending version-targeted guidance back.
+- That's the entire request + decision layer done. What's left to actually go live is the database wiring and the deploy — now being coordinated with Dawn.
+
+## Summary of New Capabilities
+
+| Capability | How to Use |
+|-----------|-----------|
+| Dispatch handlers (TS port) | `handleDispatchList` + `handleDispatchCreate` in `src/feedback-factory/dispatch/handlers.ts` — framework-agnostic, not yet wired |
+| Store dispatch seam | `FeedbackStore.listDispatches` / `findDispatchByTitle` / `createDispatch` |
+
+## Evidence
+
+- Reference is TypeScript, so equivalence is by faithful transcription plus both-sides-of-boundary tests (8): list UA-gate 400; basic list with count + asOf; version-compat filter hiding a min-version dispatch from an older agent; type filter; create 401 without the key; Bearer-form auth accepted; exact validation messages for title/content/type/version; 201-then-200-duplicate dedup by title.

package/upgrades/1.3.27.md

@@ -0,0 +1,45 @@
+# Instar Upgrade Guide — NEXT
+
+<!-- bump: patch -->
+
+## What Changed
+
+**Built-in scheduled jobs load again — the second (and final) layer of the fleet-wide
+job-load failure.** The prior fix carried the required `priority`/`expectedDurationMinutes`
+fields into each per-slug JSON manifest. That let every entry survive to the *next*
+checkpoint — the agentmd `.md` frontmatter validator — which then rejected all of them: its
+closed-set key allowlist (`ALLOWED_FRONTMATTER_KEYS`) was a "Phase 1a starter set" of
+agent-behavior keys and never included the scheduling/execution vocabulary
+(`schedule`, `priority`, `expectedDurationMinutes`, `model`, `enabled`, `tags`,
+`unrestrictedTools`, `gate`, …) that every shipped template legitimately carries in
+frontmatter. So the loader hard-rejected each built-in job on the first scheduling key it
+saw (`schedule`) with `agentmd-frontmatter-invalid`, leaving the job count at zero. The
+allowlist now includes that vocabulary. Existing `.md` files self-heal on the next load
+(their frontmatter simply stops being rejected).
+
+## What to Tell Your User
+
+- Your agent's automatic background jobs (health checks, reflection, evolution, the
+  overseers) are running again — they were quietly dead, and this revives them with nothing
+  for you to do.
+- The change only teaches the job reader to recognize scheduling labels it was wrongly
+  rejecting; it adds no new behavior, and a job still can't grant itself extra tool access
+  through those labels — the locked-down job record remains the only authority for that.
+
+## Summary of New Capabilities
+
+| Capability | How to Use |
+|-----------|-----------|
+| Built-in jobs load again | Automatic on update; `GET /jobs` shows the full set, no `agentmd-frontmatter-invalid` lines |
+
+## Evidence
+
+Found by deploying the prior fix to a real server: `jobCount` stayed `0` on v1.3.24 with 18
+fresh `agentmd-frontmatter-invalid: Unknown frontmatter key "schedule"` log lines. Proven by:
+a **gap-closer** unit test that installs every real shipped template and drives it through the
+**full** `loadAgentMdJobs` path — asserting `jobs.length === installedCount` (derived, not
+hardcoded), zero `agentmd-frontmatter-invalid`, and **valid scheduling values** (the
+assertion the prior manifest-only test missed); a closed-set regression (a genuinely-unknown
+key still rejects); a manifest-wins precedence test; a security regression (frontmatter
+`unrestrictedTools` cannot self-grant); and a live before/after on the affected server
+(`jobCount 0 → 18`). Spec converged 2 iterations across 3 reviewers.

package/upgrades/side-effects/agentmd-frontmatter-scheduling-vocabulary.md

@@ -0,0 +1,66 @@
+# Side-Effects Review — Agentmd frontmatter scheduling vocabulary (fleet-wide jobs layer 2)
+
+**Spec:** `docs/specs/AGENTMD-FRONTMATTER-SCHEDULING-VOCABULARY-SPEC.md` (converged 2 iters,
+3 reviewers; convergence report `docs/specs/reports/agentmd-frontmatter-scheduling-vocabulary-convergence.md`)
+**Change:** Expands `ALLOWED_FRONTMATTER_KEYS` in `src/scheduler/AgentMdJobLoader.ts` with the
+scheduling/execution vocabulary (`schedule, priority, expectedDurationMinutes, model, enabled,
+tags, unrestrictedTools, gate, telegramNotify, topicId, machines, supervision`). Without it, the
+loader hard-rejects every built-in job `.md` on the first scheduling key (`schedule`) →
+`jobCount=0` fleet-wide. This is layer 2 behind #425 (which fixed the JSON manifest missing the
+same fields); clearing #425's layer let entries survive to the frontmatter check, which then
+rejected them.
+**Files:** `src/scheduler/AgentMdJobLoader.ts` (allowlist + comment),
+`tests/unit/scheduler/JobLoader.agentmd.test.ts` (+3), `tests/unit/scheduler/InstallBuiltinJobs.test.ts`
+(+1 gap-closer), `tests/unit/scheduler/JobScheduler.tool-allowlist.test.ts` (+1 security).
+
+## Principle check (Phase 1)
+
+Decision point? No runtime agent-behavior gate is added or relaxed. The closed-set frontmatter
+guard is *expanded* with known keys (its own comment anticipates "the set grows … a deliberate
+change, never silent") and still rejects genuinely-unknown keys. The manifest's `validateManifest`
+remains the correctness authority; no validation is loosened.
+
+## The seven questions
+
+1. **Over-block.** Before: every built-in `.md` over-blocked (rejected on a legitimate key).
+   After: only genuinely-unknown keys reject. Correct direction.
+2. **Under-block.** The added keys are decorative — `manifestToJobDefinition` reads every
+   effective value from `manifest.*`. A malformed frontmatter copy (e.g. `priority: bogus`)
+   cannot reach a consumer because the manifest is validated independently. The closed-set is
+   regression-locked by a test (a still-unknown key must reject).
+3. **Level-of-abstraction fit.** One-file accept-list widening at the exact layer that rejects.
+   The alternative (strip keys on write) was rejected: it diverges on-disk `.md` from the shipped
+   template and breaks `frontmatterHash` lock-file verification (hashes computed over the template).
+4. **Signal vs authority.** Compliant. Manifest keeps blocking authority. Frontmatter scheduling
+   keys are inert (decorative source for the installer). No silent defaulting introduced.
+5. **Interactions.** Verified by the integration reviewer: the only scheduler read of
+   `job.frontmatter` is `resolveAllowlist` (`toolAllowlist`, already accepted pre-change, gated
+   for `*` by the MANIFEST's `unrestrictedTools`). No scheduling value is read from frontmatter.
+   Lock-file hashing unaffected (the signer always hashed the full frontmatter; the allowlist only
+   gated *reachability*, not hash inputs). **Security:** a `.md` cannot self-grant tools via
+   frontmatter — `unrestrictedTools` authority is the manifest's; a regression test pins the clamp.
+6. **External surfaces.** No new routes/config. Server-side loader code shipped via shadow-install.
+   Existing broken `.md` files self-heal on the next load (their frontmatter simply stops being
+   rejected). No agent-installed file change → **no `PostUpdateMigrator` entry required**.
+7. **Rollback cost.** Low — revert restores the bug. The added keys are inert to any reader.
+
+## Out of scope (declared)
+
+The orphan `session-reaper-promotion-review.json` on Echo's disk (`origin: custom`, null fields,
+no `.md`, no shipped template) is a **`manifest-invalid`** rejection (loader correctly rejecting an
+invalid origin), not a frontmatter issue and not fleet-wide. Swept as a one-off Echo data cleanup
+(delete the orphan manifest) so it stops emitting a per-boot warn — not part of the shipped code.
+
+## Testing
+
+- Unit (loader, +3): scheduling vocabulary accepted (no longer rejects on `schedule`); closed-set
+  preserved (genuinely-unknown key still rejects); manifest-wins precedence on disagreement.
+- Unit (installer, +1 GAP-CLOSER): install **every real shipped template** → run the **full**
+  `loadAgentMdJobs` → assert `jobs.length === installedCount` (dynamic, not hardcoded), **zero**
+  `agentmd-frontmatter-invalid`, and **valid scheduling VALUES** (non-empty schedule, valid
+  priority, duration > 0). This is the end-to-end assertion #425's manifest-only test missed.
+- Unit (scheduler, +1 SECURITY): frontmatter `unrestrictedTools:true` + `toolAllowlist:'*'` with
+  manifest `unrestrictedTools:false` → clamps to `['Read']` (no self-grant).
+- **Live evidence (test-as-self):** `loadAgentMdJobs` over Echo's real jobs dir — unpatched 0 →
+  patched 18. Full live check (restart Echo, `GET /jobs` 0→18, no new `agentmd-frontmatter-invalid`)
+  to be captured post-publish before declaring fixed (bug-fix evidence bar).