diff --git a/ci-operator/step-registry/openshift/claude/agent-eval/openshift-claude-agent-eval-commands.sh b/ci-operator/step-registry/openshift/claude/agent-eval/openshift-claude-agent-eval-commands.sh
index 0851c71909fad..ac266a5fcdeb7 100755
--- a/ci-operator/step-registry/openshift/claude/agent-eval/openshift-claude-agent-eval-commands.sh
+++ b/ci-operator/step-registry/openshift/claude/agent-eval/openshift-claude-agent-eval-commands.sh
@@ -7,6 +7,7 @@
 #
 # Optional env:
 #   EVAL_MODEL        -- model for the skill under test (default: claude-sonnet-4-6)
+#   EVAL_EFFORT       -- agent reasoning effort passed to /eval-run as --effort
 #   EVAL_PARALLELISM  -- number of test cases to run concurrently (default: 1)
 #   EVAL_CASES        -- comma-separated list of case IDs to run (default: all)
 #   EVAL_DISCOVER     -- "true" or glob pattern to auto-discover eval configs
@@ -27,6 +28,10 @@ if [[ -n "${MULTISTAGE_PARAM_OVERRIDE_EVAL_MODEL:-}" ]]; then
     echo "Applying Gangway override: EVAL_MODEL=${MULTISTAGE_PARAM_OVERRIDE_EVAL_MODEL}"
     EVAL_MODEL="${MULTISTAGE_PARAM_OVERRIDE_EVAL_MODEL}"
 fi
+if [[ -n "${MULTISTAGE_PARAM_OVERRIDE_EVAL_EFFORT:-}" ]]; then
+    echo "Applying Gangway override: EVAL_EFFORT=${MULTISTAGE_PARAM_OVERRIDE_EVAL_EFFORT}"
+    EVAL_EFFORT="${MULTISTAGE_PARAM_OVERRIDE_EVAL_EFFORT}"
+fi
 
 # Load GitHub token for gh CLI access (same secret as payload-agent)
 set +x
@@ -498,6 +503,7 @@ for config in "${CONFIGS_TO_RUN[@]}"; do
     fi
 
     EVAL_RUN_ARGS="--config ${config} --model ${EVAL_MODEL} --run-id ${RUN_ID} --parallelism ${EVAL_PARALLELISM}"
+    [[ -n "${EVAL_EFFORT:-}" ]] && EVAL_RUN_ARGS="${EVAL_RUN_ARGS} --effort ${EVAL_EFFORT}"
     [[ -n "${CASE_ARGS}" ]] && EVAL_RUN_ARGS="${EVAL_RUN_ARGS} ${CASE_ARGS}"
     [[ -n "${EVAL_BASELINE}" ]] && EVAL_RUN_ARGS="${EVAL_RUN_ARGS} --baseline ${EVAL_BASELINE}"
     [[ -n "${EVAL_EXTRA_ARGS}" ]] && EVAL_RUN_ARGS="${EVAL_RUN_ARGS} ${EVAL_EXTRA_ARGS}"
diff --git a/ci-operator/step-registry/openshift/claude/agent-eval/openshift-claude-agent-eval-ref.yaml b/ci-operator/step-registry/openshift/claude/agent-eval/openshift-claude-agent-eval-ref.yaml
index 7868e87b9d868..1e218fe2b499a 100644
--- a/ci-operator/step-registry/openshift/claude/agent-eval/openshift-claude-agent-eval-ref.yaml
+++ b/ci-operator/step-registry/openshift/claude/agent-eval/openshift-claude-agent-eval-ref.yaml
@@ -19,6 +19,18 @@ ref:
     documentation: |-
       Gangway API override for EVAL_MODEL. When triggering via
       Gangway, pass as pod_spec_options.envs.
+  - name: EVAL_EFFORT
+    default: ""
+    documentation: |-
+      Agent reasoning effort passed to /eval-run as --effort (e.g.
+      low, medium, high, xhigh for the codex runner; the claude-code
+      runner accepts its own levels). Empty means use the runner.effort
+      default from the eval config.
+  - name: MULTISTAGE_PARAM_OVERRIDE_EVAL_EFFORT
+    default: ""
+    documentation: |-
+      Gangway API override for EVAL_EFFORT. When triggering via
+      Gangway, pass as pod_spec_options.envs.
   - name: EVAL_BASELINE
     default: ""
   - name: EVAL_PARALLELISM