diff --git a/src/lib/onboard/machine/flow-phases/agent-policy-finalization.test.ts b/src/lib/onboard/machine/flow-phases/agent-policy-finalization.test.ts
new file mode 100644
index 0000000000..749b57c4a2
--- /dev/null
+++ b/src/lib/onboard/machine/flow-phases/agent-policy-finalization.test.ts
@@ -0,0 +1,182 @@
+// SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+import { describe, expect, it } from "vitest";
+
+import {
+  createSession,
+  filterSafeUpdates,
+  MACHINE_SNAPSHOT_VERSION,
+  normalizeSession,
+  sanitizeFailure,
+  type Session,
+  type SessionUpdates,
+} from "../../../state/onboard-session";
+import type { OnboardFlowContext } from "../flow-context";
+import { advanceTo, completeOnboardMachine } from "../result";
+import { OnboardRuntime, type OnboardRuntimeDeps } from "../runtime";
+import { runOnboardSequenceWithRunner } from "../sequence-runner";
+import {
+  createAgentSetupPhase,
+  createFinalizationPhase,
+  createOpenclawSetupPhase,
+  createPoliciesPhase,
+  createPostVerifyPhase,
+} from "./agent-policy-finalization";
+
+function context(): OnboardFlowContext<null, null, null> {
+  return {
+    resume: false,
+    fresh: false,
+    session: createSession(),
+    agent: null,
+    recordedSandboxName: null,
+    requestedSandboxName: null,
+    sandboxName: "my-assistant",
+    fromDockerfile: null,
+    model: "model",
+    provider: "provider",
+    endpointUrl: null,
+    credentialEnv: null,
+    hermesAuthMethod: null,
+    hermesToolGateways: [],
+    preferredInferenceApi: null,
+    nimContainer: null,
+    webSearchConfig: null,
+    webSearchSupported: true,
+    selectedMessagingChannels: [],
+    gpu: null,
+    sandboxGpuConfig: null,
+    gpuPassthrough: false,
+  };
+}
+
+function cloneSession(session: Session): Session {
+  return normalizeSession(JSON.parse(JSON.stringify(session))) ?? session;
+}
+
+function createRuntime(initialSession: Session = createSession()) {
+  let session = cloneSession(initialSession);
+  const updateSession = (mutator: (value: Session) => Session | void): Session => {
+    session = cloneSession(mutator(cloneSession(session)) ?? session);
+    return cloneSession(session);
+  };
+  const deps: OnboardRuntimeDeps = {
+    loadSession: () => cloneSession(session),
+    createSession,
+    saveSession: (next) => {
+      session = cloneSession(next);
+      return cloneSession(session);
+    },
+    updateSession,
+    markStepStarted: () => cloneSession(session),
+    markStepComplete: (_stepName, updates: SessionUpdates = {}) =>
+      updateSession((current) => {
+        Object.assign(current, filterSafeUpdates(updates));
+        return current;
+      }),
+    markStepCompleteRecordOnly: (_stepName, updates: SessionUpdates = {}) =>
+      updateSession((current) => {
+        Object.assign(current, filterSafeUpdates(updates));
+        return current;
+      }),
+    markStepSkipped: () => cloneSession(session),
+    markStepFailed: (stepName, message) =>
+      updateSession((current) => {
+        current.status = "failed";
+        current.failure = sanitizeFailure({ step: stepName, message, recordedAt: "now" });
+        return current;
+      }),
+    markStepFailedRecordOnly: () => cloneSession(session),
+    completeSession: (updates: SessionUpdates = {}) =>
+      updateSession((current) => {
+        Object.assign(current, filterSafeUpdates(updates));
+        current.status = "complete";
+        current.resumable = false;
+        return current;
+      }),
+    filterSafeUpdates,
+    emitEvent: () => undefined,
+    now: () => "2026-05-29T00:00:00.000Z",
+  };
+  return new OnboardRuntime(deps);
+}
+
+describe("agent/policy/finalization phases", () => {
+  it("creates branch-specific setup phases", async () => {
+    const agentPhase = createAgentSetupPhase(async () => ({ result: advanceTo("policies") }));
+    const openclawPhase = createOpenclawSetupPhase(async () => ({ result: advanceTo("policies") }));
+
+    expect(agentPhase.state).toBe("agent_setup");
+    expect(openclawPhase.state).toBe("openclaw");
+    await expect(agentPhase.run(context())).resolves.toMatchObject({
+      result: { next: "policies" },
+    });
+    await expect(openclawPhase.run(context())).resolves.toMatchObject({
+      result: { next: "policies" },
+    });
+  });
+
+  it("maps policies context updates", async () => {
+    const phase = createPoliciesPhase(async () => ({
+      context: { selectedMessagingChannels: ["slack"] },
+      result: advanceTo("finalizing"),
+    }));
+
+    const result = await phase.run(context());
+
+    expect(phase.state).toBe("policies");
+    expect(result.context.selectedMessagingChannels).toEqual(["slack"]);
+    expect(result.result).toMatchObject({ next: "finalizing" });
+  });
+
+  it("creates finalization and post-verify phases", async () => {
+    const finalizing = createFinalizationPhase(async () => ({ result: advanceTo("post_verify") }));
+    const postVerify = createPostVerifyPhase(async () => ({
+      result: completeOnboardMachine({ sandboxName: "my-assistant" }),
+    }));
+
+    expect(finalizing.state).toBe("finalizing");
+    expect(postVerify.state).toBe("post_verify");
+    await expect(finalizing.run(context())).resolves.toMatchObject({
+      result: { next: "post_verify" },
+    });
+    await expect(postVerify.run(context())).resolves.toMatchObject({
+      result: { type: "complete" },
+    });
+  });
+
+  it("runs branch-to-completion phases through the strict FSM runner", async () => {
+    const initialSession = createSession({
+      machine: {
+        version: MACHINE_SNAPSHOT_VERSION,
+        state: "openclaw",
+        stateEnteredAt: "2026-05-29T00:00:00.000Z",
+        revision: 0,
+      },
+    });
+
+    const result = await runOnboardSequenceWithRunner({
+      context: context(),
+      runtime: createRuntime(initialSession),
+      phases: [
+        createOpenclawSetupPhase(async () => ({ result: advanceTo("policies") })),
+        createPoliciesPhase(async () => ({
+          context: { selectedMessagingChannels: ["slack"] },
+          result: advanceTo("finalizing"),
+        })),
+        createFinalizationPhase(async () => ({ result: advanceTo("post_verify") })),
+        createPostVerifyPhase(async () => ({
+          result: completeOnboardMachine({ sandboxName: "my-assistant" }),
+        })),
+      ],
+    });
+
+    expect(result.session).toMatchObject({
+      status: "complete",
+      sandboxName: "my-assistant",
+      machine: { state: "complete" },
+    });
+    expect(result.context.selectedMessagingChannels).toEqual(["slack"]);
+  });
+});
diff --git a/src/lib/onboard/machine/flow-phases/agent-policy-finalization.ts b/src/lib/onboard/machine/flow-phases/agent-policy-finalization.ts
new file mode 100644
index 0000000000..95ff78a95f
--- /dev/null
+++ b/src/lib/onboard/machine/flow-phases/agent-policy-finalization.ts
@@ -0,0 +1,57 @@
+// SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+import type { OnboardFlowContext, OnboardFlowPhaseResult } from "../flow-context";
+import { mergeOnboardFlowContext, onboardFlowPhaseResult } from "../flow-context";
+import type { OnboardSequencePhase } from "../sequence-runner";
+
+type FlowPhaseHandler<Context extends OnboardFlowContext> = (context: Context) => Promise<{
+  context?: Partial<Context>;
+  result: OnboardFlowPhaseResult<Context>["result"];
+}>;
+
+function createFlowPhase<Context extends OnboardFlowContext>(
+  state: OnboardSequencePhase<Context>["state"],
+  runPhase: FlowPhaseHandler<Context>,
+): OnboardSequencePhase<Context> {
+  return {
+    state,
+    async run(context) {
+      const result = await runPhase(context);
+      return onboardFlowPhaseResult(
+        result.context ? mergeOnboardFlowContext(context, result.context) : context,
+        result.result,
+      );
+    },
+  };
+}
+
+export function createAgentSetupPhase<Context extends OnboardFlowContext>(
+  runAgentSetup: FlowPhaseHandler<Context>,
+): OnboardSequencePhase<Context> {
+  return createFlowPhase("agent_setup", runAgentSetup);
+}
+
+export function createOpenclawSetupPhase<Context extends OnboardFlowContext>(
+  runOpenclawSetup: FlowPhaseHandler<Context>,
+): OnboardSequencePhase<Context> {
+  return createFlowPhase("openclaw", runOpenclawSetup);
+}
+
+export function createPoliciesPhase<Context extends OnboardFlowContext>(
+  runPolicies: FlowPhaseHandler<Context>,
+): OnboardSequencePhase<Context> {
+  return createFlowPhase("policies", runPolicies);
+}
+
+export function createFinalizationPhase<Context extends OnboardFlowContext>(
+  runFinalization: FlowPhaseHandler<Context>,
+): OnboardSequencePhase<Context> {
+  return createFlowPhase("finalizing", runFinalization);
+}
+
+export function createPostVerifyPhase<Context extends OnboardFlowContext>(
+  runPostVerify: FlowPhaseHandler<Context>,
+): OnboardSequencePhase<Context> {
+  return createFlowPhase("post_verify", runPostVerify);
+}