Regression: Workflow with skipped step stuck because failed to resolve references

[spaced]

Pre-requisites

• I have double-checked my configuration
• I have tested with the :latest image tag (i.e. quay.io/argoproj/workflow-controller:latest) and can confirm the issue still exists on :latest. If not, I have explained why, in detail, in my description below.
• I have searched existing issues and could not find a match for this bug
• I'd like to contribute the fix myself (see contributing guide)

What happened? What did you expect to happen?

works until v4.0.1, but not with > v4.0.2

when a step is skipped,
but next step has an expression like

"{{= steps['sayhiskippy'].status == 'Skipped' ? 'skippy' : steps['sayhiskippy'].outputs.parameters.out }}"

Version(s)

v4.0.2, v4.0.3,v4.0.4,v4.0.5

Paste a minimal workflow that reproduces the issue. We must be able to run the workflow; don't enter a workflow that uses private images.

---
apiVersion: argoproj.io/v1alpha1
kind: WorkflowTemplate
metadata:
  name: test-skipped-step
spec:
  entrypoint: main
  arguments:
    parameters:
      - name: a_string
        value: 'one-two'
  templates:
    - name: main
      steps:
        - - name: sayhiskippy
            template: sayhi
            arguments:
              parameters:
                - name: text
                  value: "this is the skippy"
            when: "{{= workflow.parameters.a_string != 'one-two' }}"
        - - name: afterskippy
            template: sayhi
            arguments:
              parameters:
                - name: text
                  value: "{{= steps['sayhiskippy'].status == 'Skipped' ? 'skippy' : steps['sayhiskippy'].outputs.parameters.out }}"

    - name: sayhi
      inputs:
        parameters:
          - name: text
      outputs:
        parameters:
          - name: out
            valueFrom:
              path: /tmp/out
      script:
        image: alpine:latest
        command:
          - sh
        source: |
          echo "hello {{inputs.parameters.text }}" > /tmp/out

Logs from the workflow controller

kubectl logs -n argo deploy/workflow-controller | grep ${workflow}

{"time":"2026-05-20T14:23:08.023039084Z","level":"WARN","msg":"was unable to find variable","workflow":"test-skipped-step-2ktbs","namespace":"**","error":"failed to evaluate expression
: steps.sayhiskippy.outputs is missing","component":"workflow_worker"}
{"time":"2026-05-20T14:23:08.023059164Z","level":"ERROR","msg":"Failed to resolve references","component":"workflow_worker","workflow":"test-skipped-step-2ktbs","error":"requeue","namespace"
:"**","stepName":"afterskippy"}

Logs from in your workflow's wait container

kubectl logs -n argo -c wait -l workflows.argoproj.io/workflow=${workflow},workflow.argoproj.io/phase!=Succeeded

[coderabbitai]

⚠️ Possible Duplicate Issue(s)

• Workflow hangs when 'when' clause references skipped step outputs #15873

🔗 Related PRs

#15736 - fix: add stepgroup and taskgroup to scope. Fixes #15737 [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[spaced]

added test to main branch commit 7cce696e0 , fails with

time=2026-05-20T19:52:13.800+02:00 level=WARN msg="was unable to find variable" error="failed to evaluate expression: steps.A.outputs is missing" test=true workflow="" namespace=""
time=2026-05-20T19:52:13.800+02:00 level=ERROR msg="Failed to resolve references" test=true workflow="" namespace="" stepName=B error=requeue
time=2026-05-20T19:52:13.800+02:00 level=INFO msg="Workflow step group node not yet completed" workflow="" namespace="" nodeID=-1981202788 test=true
time=2026-05-20T19:52:13.800+02:00 level=INFO msg="TaskSet Reconciliation" test=true workflow="" namespace=""
time=2026-05-20T19:52:13.800+02:00 level=INFO msg=reconcileAgentPod workflow="" namespace="" test=true
time=2026-05-20T19:52:13.800+02:00 level=INFO msg="Workflow update successful" test=true workflow="" namespace="" resourceVersion="" phase=Running lastSeenVersion=""
    steps_test.go:788: 
        	Error Trace:	/home/**/git/argo-workflows/workflow/controller/steps_test.go:788
        	Error:      	Expected value not to be nil.
        	Test:       	TestStepsWhenSkipExpression
--- FAIL: TestStepsWhenSkipExpression (0.13s)

Index: workflow/controller/steps_test.go
===================================================================
diff --git a/workflow/controller/steps_test.go b/workflow/controller/steps_test.go
--- a/workflow/controller/steps_test.go	(revision 85c2103770c28522c4ef030b861656f66841ea39)
+++ b/workflow/controller/steps_test.go	(date 1779289177485)
@@ -736,6 +736,63 @@
 	assert.Equal(t, wfv1.NodeSkipped, nodeB.Phase)
 }
 
+var stepsWhenSkipWithExpression = `
+apiVersion: argoproj.io/v1alpha1
+kind: Workflow
+metadata:
+  generateName: steps-when-skip-no-requeue-
+spec:
+  entrypoint: main
+  templates:
+  - name: main
+    steps:
+    - - name: A
+        template: script-echo
+        when: "false"
+    - - name: B
+        template: echo-with-param
+        arguments:
+          parameters:
+          - name: msg
+            value: "{{= steps.A.status == 'Skipped' ? 'skipped' : steps.A.outputs.result}}"
+  - name: script-echo
+    script:
+      image: alpine:3.23
+      command: [sh]
+      source: |
+        echo hello
+  - name: echo-with-param
+    inputs:
+      parameters:
+      - name: msg
+    container:
+      image: alpine:3.23
+      command: [echo, "{{inputs.parameters.msg}}"]
+`
+
+func TestStepsWhenSkipExpression(t *testing.T) {
+	ctx := logging.TestContext(t.Context())
+	cancel, controller := newController(ctx)
+	defer cancel()
+	wfcset := controller.wfclientset.ArgoprojV1alpha1().Workflows("")
+
+	wf := wfv1.MustUnmarshalWorkflow(stepsWhenSkipWithExpression)
+	wf, err := wfcset.Create(ctx, wf, metav1.CreateOptions{})
+	require.NoError(t, err)
+	woc := newWorkflowOperationCtx(ctx, wf, controller)
+
+	woc.operate(ctx)
+
+	// Verify the resolved value of steps.B.msg
+	nodeB := woc.wf.Status.Nodes.FindByDisplayName("B")
+	require.NotNil(t, nodeB)
+
+	require.NotNil(t, nodeB.Inputs)
+	require.Len(t, nodeB.Inputs.Parameters, 1)
+	assert.Equal(t, "skipped", nodeB.Inputs.Parameters[0].Value.String())
+	assert.Equal(t, wfv1.WorkflowRunning, woc.wf.Status.Phase)
+}
+
 var stepsWhenExprWithParamFilter = `
 apiVersion: argoproj.io/v1alpha1
 kind: Workflow

what i also see, ugly WARN logs like

time=2026-05-20T16:59:41.080+02:00 level=WARN msg="Node was nil, will be initialized as type Skipped" test=true workflow="" namespace=""

can we get rid of this?