diff --git a/services/hermes/switchyard-configmap.yaml b/services/hermes/switchyard-configmap.yaml
index 449e9c56..b1052518 100644
--- a/services/hermes/switchyard-configmap.yaml
+++ b/services/hermes/switchyard-configmap.yaml
@@ -1098,6 +1098,9 @@ data:
       work.
     - low: only mechanical, reversible, tightly bounded work such as a typo,
       formatting, or a simple lookup.
+    An implementation or test change requires at least medium effort even when
+    it is small or well-scoped; low is reserved for mechanical edits and
+    lookups. Keep this effort floor independent from capability selection.
     Never choose below the floor and never exceed xhigh.
     If the objective records a failing test, failed attempt, contradicted
     result, rejected review, or incomplete evidence from a previous worker,
@@ -1131,6 +1134,7 @@ data:
     Implement critical security review fixes -> worker_codex_auto_advanced_xhigh.
     Difficult intermittent production failure -> worker_codex_auto_advanced_high.
     Ordinary component design -> worker_claude_auto_balanced_medium.
+    Small implementation with focused tests -> worker_codex_auto_balanced_medium.
     One spelling correction -> worker_codex_auto_economy_low.
     Anthropic exhausted + critical independent final review ->
     worker_codex_auto_frontier_xhigh.
diff --git a/services/hermes/switchyard-deployment.yaml b/services/hermes/switchyard-deployment.yaml
index 88f29b9b..b3c55d0f 100644
--- a/services/hermes/switchyard-deployment.yaml
+++ b/services/hermes/switchyard-deployment.yaml
@@ -22,7 +22,7 @@ spec:
       labels:
         app: hermes-switchyard
       annotations:
-        ai.bstein.dev/config-rev: "20260913-capability-effort-v4"
+        ai.bstein.dev/config-rev: "20260913-capability-effort-v5"
         prometheus.io/scrape: "true"
         prometheus.io/port: "9005"
         prometheus.io/path: /metrics