diff --git a/packages/dashboard/app/utils/__tests__/taskProgress.test.ts b/packages/dashboard/app/utils/__tests__/taskProgress.test.ts
index 938e87f3de..10a5a6b201 100644
--- a/packages/dashboard/app/utils/__tests__/taskProgress.test.ts
+++ b/packages/dashboard/app/utils/__tests__/taskProgress.test.ts
@@ -5,7 +5,7 @@ import { getUnifiedTaskProgress } from "../taskProgress";
 /*
 FNXC:WorkflowSteps 2026-06-25-00:00 — graph-native progress model (plan U3).
 These tests pin the render-state contract that the progress bar / Workflow tab rely on:
-- names resolve from result.workflowStepName (no DB-row name lookup), with a raw-id fallback;
+- names resolve from result.workflowStepName (no DB-row name lookup), with a humanized node-id fallback;
 - a "pending" result with a startedAt and no completedAt is the `running` state, vs bare `pending`;
 - advisory_failure (non-blocking) is distinct from failed (blocking) and counts as completed;
 - disabled optional steps (absent from enabledWorkflowSteps) never appear in the counter/bar.
@@ -49,11 +49,25 @@ describe("getUnifiedTaskProgress", () => {
     );
 
     const item = progress.items.find((i) => i.id === "workflow-code-review");
-    expect(item?.name).toBe("code-review");
+    // Enabled-but-not-run has no recorded name → humanize the node id to proper casing,
+    // never render the raw lowercase id.
+    expect(item?.name).toBe("Code Review");
     // Enabled but never run → pending.
     expect(item?.status).toBe("pending");
   });
 
+  it("humanizes the node id to proper casing for an enabled-but-not-run step", () => {
+    const progress = getUnifiedTaskProgress(
+      makeTask({
+        enabledWorkflowSteps: ["browser-verification", "frontend-ux-design", "code-review"],
+        workflowStepResults: [],
+      }),
+    );
+    expect(progress.items.find((i) => i.id === "workflow-browser-verification")?.name).toBe("Browser Verification");
+    expect(progress.items.find((i) => i.id === "workflow-frontend-ux-design")?.name).toBe("Frontend UX Design");
+    expect(progress.items.find((i) => i.id === "workflow-code-review")?.name).toBe("Code Review");
+  });
+
   it("distinguishes running (started, not completed) from pending (not started)", () => {
     const progress = getUnifiedTaskProgress(
       makeTask({
diff --git a/packages/dashboard/app/utils/taskProgress.ts b/packages/dashboard/app/utils/taskProgress.ts
index 173c00410f..71f249c30c 100644
--- a/packages/dashboard/app/utils/taskProgress.ts
+++ b/packages/dashboard/app/utils/taskProgress.ts
@@ -59,12 +59,32 @@ function isCompleted(status: UnifiedTaskProgressStatus): boolean {
   return status === "done" || status === "skipped" || status === "advisory_failure";
 }
 
+/*
+FNXC:WorkflowStepResults 2026-06-26-16:30:
+An enabled-but-not-yet-run workflow step has no recorded result yet, so there is no
+`workflowStepName` to show. Rather than render the raw graph node id (e.g. `code-review`,
+`browser-verification`), humanize it into a Title Case label ("Code Review",
+"Browser Verification"). Once the graph records the step it carries the workflow's exact
+`config.name`, which always wins; humanization is only the pre-run fallback. The UI must
+show proper casing for workflow steps (e.g. "Code Review"), never the lowercase hyphenated id.
+*/
+function humanizeWorkflowStepId(workflowStepId: string): string {
+  const words = workflowStepId
+    .replace(/^plugin:/, "")
+    .split(/[-_:\s]+/)
+    .filter(Boolean);
+  if (words.length === 0) return workflowStepId;
+  return words
+    .map((w) => (/^(ux|ui|qa|ai|api|pr|id)$/i.test(w) ? w.toUpperCase() : w.charAt(0).toUpperCase() + w.slice(1)))
+    .join(" ");
+}
+
 function resolveWorkflowStepName(workflowStepId: string, result: WorkflowStepResult | undefined): string {
   const resultName = result?.workflowStepName?.trim();
   if (resultName) {
     return resultName;
   }
-  return workflowStepId;
+  return humanizeWorkflowStepId(workflowStepId);
 }
 
 export function getUnifiedTaskProgress(