diff --git a/behaviors/core-expert.yaml b/behaviors/core-expert.yaml
index 95c5b147..53768849 100644
--- a/behaviors/core-expert.yaml
+++ b/behaviors/core-expert.yaml
@@ -1,17 +1,19 @@
 bundle:
   name: core-expert-behavior
-  version: 1.0.0
+  version: 1.1.0
   description: Expert consultant for Amplifier kernel internals
 
-context:
-  include:
-    - core:context/release-mandate.md
-
 agents:
   include:
     - core:core-expert
 
-# Note: kernel-overview.md context moved to core-expert.md agent
-# This reduces root session token usage - context loads only when agent spawns
-# Exception: release-mandate.md is in context.include because it MUST be visible
-# in every root session working on amplifier-core, not just when core-expert spawns.
+# Both kernel-overview.md and release-mandate.md are loaded via the core-expert
+# agent body (@-mentioned). Always-on context.include is intentionally empty —
+# the agent's meta.description is the discovery mechanism for "release mandate /
+# kernel knowledge exists; delegate to core-expert for release work."
+#
+# Prior versions kept release-mandate.md in always-on under the reasoning that
+# every root session working on amplifier-core needed it. Re-evaluated against
+# foundation's context-sink discipline: the doc is ~2.5K tokens, only relevant
+# when actually cutting a release, and the core-expert agent is the canonical
+# delegation target for release work. Always-on cost was unjustified.
diff --git a/bundle.md b/bundle.md
index 4b929633..784e37f7 100644
--- a/bundle.md
+++ b/bundle.md
@@ -1,7 +1,7 @@
 ---
 bundle:
   name: core
-  version: 1.0.0
+  version: 1.1.0
   description: Amplifier kernel - ultra-thin mechanism layer providing stable contracts
 
 includes: