Harden world-class release gates

examples/governed-incident-command/optimization/reports/description_optimization.md

@@ -14,18 +14,18 @@ Build governed incident command packets. Use when asked to standardize incident
 
 | Candidate | Tokens | Dev FP | Dev FN | Dev Near | Holdout FP | Holdout FN |
 | --- | ---: | ---: | ---: | ---: | ---: | ---: |
-| `Current` | 37 | 0 | 1 | 1.0 | 0 | 1 |
-| `Guardrail` | 51 | 0 | 1 | 1.0 | 1 | 1 |
-| `Balanced` | 54 | 0 | 1 | 1.0 | 0 | 1 |
-| `Boundary` | 78 | 0 | 1 | 1.0 | 0 | 1 |
-| `Artifact Aware` | 78 | 0 | 1 | 1.0 | 0 | 1 |
+| `Current` | 37 | 0 | 0 | 1.0 | 0 | 1 |
+| `Balanced` | 54 | 0 | 0 | 1.0 | 0 | 1 |
+| `Boundary` | 78 | 0 | 0 | 1.0 | 0 | 1 |
+| `Artifact Aware` | 78 | 0 | 0 | 1.0 | 0 | 1 |
+| `Guardrail` | 51 | 0 | 1 | 1.0 | 0 | 1 |
 
 ## Acceptance Gates
 
 | Gate | Winner FP | Winner FN | Current FP | Current FN | Baseline FP | Baseline FN |
 | --- | ---: | ---: | ---: | ---: | ---: | ---: |
-| Holdout | 0 | 1 | 0 | 1 | 0 | 2 |
-| Blind Holdout | 0 | 0 | 0 | 0 | 0 | 1 |
+| Holdout | 0 | 1 | 0 | 1 | 0 | 1 |
+| Blind Holdout | 0 | 0 | 0 | 0 | 0 | 0 |
 | Judge Blind Holdout | 0 | 0 | 0 | 0 | 0 | 0 |
 | Adversarial Holdout | 0 | 0 | 0 | 0 | 0 | 0 |
 
@@ -34,21 +34,21 @@ Build governed incident command packets. Use when asked to standardize incident
 | Gate | Winner Gap | Winner Risk | Winner Boundary Rate | Current Gap | Baseline Gap |
 | --- | ---: | --- | ---: | ---: | ---: |
 | Holdout | 0.023 | overlap | 0.0 | 0.023 | 0.101 |
-| Blind Holdout | 0.418 | healthy | 0.167 | 0.418 | 0.063 |
-| Adversarial Holdout | 0.598 | healthy | 0.0 | 0.598 | 0.172 |
+| Blind Holdout | 0.417 | healthy | 0.0 | 0.417 | 0.374 |
+| Adversarial Holdout | 0.538 | watch | 0.167 | 0.538 | 0.467 |
 
 ## Judge Blind Summary
 
 | Gate | Winner Agreement | Winner Mean Confidence | Current Agreement | Baseline Agreement |
 | --- | ---: | ---: | ---: | ---: |
-| Judge Blind Holdout | 1.0 | 0.657 | 1.0 | 1.0 |
+| Judge Blind Holdout | 1.0 | 0.677 | 1.0 | 1.0 |
 
 ## Family Health
 
 | Gate | Winner Clean Families | Winner Weakest Family | Current Clean Families | Baseline Clean Families |
 | --- | --- | --- | --- | --- |
-| Holdout | 5/6 | packet_assembly (1 errors) | 5/6 | 4/6 |
-| Blind Holdout | 6/6 | blind_summary_only (0 errors) | 6/6 | 5/6 |
+| Holdout | 5/6 | packet_assembly (1 errors) | 5/6 | 5/6 |
+| Blind Holdout | 6/6 | blind_summary_only (0 errors) | 6/6 | 6/6 |
 | Judge Blind Holdout | 6/6 | blind_summary_only (0 errors) | 6/6 | 6/6 |
 | Adversarial Holdout | 6/6 | adversarial_single_update_collision (0 errors) | 6/6 | 6/6 |
 

examples/governed-incident-command/optimization/semantic_config.json

@@ -37,9 +37,12 @@
       "weight": 0.32,
       "phrases": [
         "incident command packet",
+        "incident command packets",
         "governed incident packet",
         "incident packet assembly",
-        "incident review skill"
+        "incident review skill",
+        "incident review",
+        "standardize incident review"
       ]
     },
     "incident_inputs": {
@@ -59,6 +62,8 @@
         "severity assessment",
         "severity",
         "impact scope",
+        "incident review",
+        "outage review",
         "sev"
       ]
     },
@@ -68,6 +73,8 @@
         "stakeholder update",
         "incident communication",
         "owner",
+        "owners",
+        "owner actions",
         "action ownership",
         "next actions"
       ]
@@ -76,6 +83,9 @@
       "weight": 0.1,
       "phrases": [
         "governed",
+        "governed workflow",
+        "reusable governed workflow",
+        "standardize",
         "reviewed",
         "audited",
         "maintained",
@@ -128,9 +138,13 @@
       "exclusive": true,
       "phrases": [
         "brainstorm",
+        "sketch a future",
+        "for later",
         "possible response ideas",
         "improve incident handling",
-        "ideas for later"
+        "ideas for later",
+        "do not package",
+        "do not package a reusable workflow"
       ]
     },
     "summary_only": {