diff --git a/src/lib/actions/sandbox/rebuild-messaging-stage.test.ts b/src/lib/actions/sandbox/rebuild-messaging-stage.test.ts
index b5f0649476b..118f832aea3 100644
--- a/src/lib/actions/sandbox/rebuild-messaging-stage.test.ts
+++ b/src/lib/actions/sandbox/rebuild-messaging-stage.test.ts
@@ -30,6 +30,21 @@ const { stageMessagingManifestPlanForRebuild } = D("actions/sandbox/rebuild.js")
   ) => Promise<unknown>;
 };
 
+const emptyStoredMessagingPlan = {
+  schemaVersion: 1,
+  sandboxName: "openclaw-sandbox",
+  agent: "openclaw",
+  workflow: "remove-channel",
+  channels: [],
+  disabledChannels: [],
+  credentialBindings: [],
+  networkPolicy: { presets: [], entries: [] },
+  agentRender: [],
+  buildSteps: [],
+  stateUpdates: [],
+  healthChecks: [],
+};
+
 describe("stageMessagingManifestPlanForRebuild non-messaging agent guard", () => {
   afterEach(() => {
     vi.restoreAllMocks();
@@ -61,6 +76,38 @@ describe("stageMessagingManifestPlanForRebuild non-messaging agent guard", () =>
     expect(result).toBeNull();
   });
 
+  it("stages an explicit empty rebuild plan so token-backed channels are not rediscovered", async () => {
+    vi.spyOn(defs, "loadAgent").mockReturnValue({
+      name: "openclaw",
+      messagingPlatforms: ["telegram"],
+    });
+    const clearPlanEnvSpy = vi.spyOn(messaging.MessagingSetupApplier, "clearPlanEnv");
+    const writePlanEnvSpy = vi
+      .spyOn(messaging.MessagingSetupApplier, "writePlanToEnv")
+      .mockImplementation(() => undefined);
+
+    const messages: string[] = [];
+    const result = await stageMessagingManifestPlanForRebuild(
+      "openclaw-sandbox",
+      {
+        name: "openclaw-sandbox",
+        messaging: { schemaVersion: 1, plan: emptyStoredMessagingPlan },
+      },
+      "openclaw",
+      (msg) => messages.push(msg),
+    );
+
+    expect(clearPlanEnvSpy).not.toHaveBeenCalled();
+    expect(writePlanEnvSpy).toHaveBeenCalledWith(
+      expect.objectContaining({
+        workflow: "rebuild",
+        channels: [],
+      }),
+    );
+    expect(messages).toContain("Messaging manifest rebuild plan staged: no configured channels");
+    expect(result).toMatchObject({ workflow: "rebuild", channels: [] });
+  });
+
   it("emits the empty-allowlist skip message for a known agent whose messagingPlatforms is an explicit empty allowlist", async () => {
     vi.spyOn(defs, "loadAgent").mockReturnValue({
       name: "openclaw",
diff --git a/src/lib/messaging/compiler/workflow-planner.test.ts b/src/lib/messaging/compiler/workflow-planner.test.ts
index 06271511528..6d18f666c6b 100644
--- a/src/lib/messaging/compiler/workflow-planner.test.ts
+++ b/src/lib/messaging/compiler/workflow-planner.test.ts
@@ -8,7 +8,7 @@ import {
   createBuiltInRenderTemplateResolver,
 } from "../channels";
 import { createBuiltInMessagingHookRegistry, MessagingHookRegistry } from "../hooks";
-import { createChannelManifestRegistry, type ChannelManifest } from "../manifest";
+import { type ChannelManifest, createChannelManifestRegistry } from "../manifest";
 import { MessagingWorkflowPlanner } from "./workflow-planner";
 
 const TEST_CREDENTIALS: Readonly<Record<string, string>> = {
@@ -785,6 +785,47 @@ describe("MessagingWorkflowPlanner", () => {
     expect(removed?.agentRender.some((entry) => entry.channelId === "telegram")).toBe(false);
   });
 
+  it("preserves an explicit empty plan on rebuild after the final channel is removed", async () => {
+    const existingPlan = await planner().buildPlan({
+      sandboxName: "demo",
+      agent: "openclaw",
+      workflow: "onboard",
+      isInteractive: false,
+      configuredChannels: ["telegram"],
+      credentialAvailability: { TELEGRAM_BOT_TOKEN: true },
+    });
+
+    const removed = await planner().buildChannelRemovePlanFromSandboxEntry({
+      sandboxName: "demo",
+      agent: "openclaw",
+      sandboxEntry: {
+        name: "demo",
+        messaging: { schemaVersion: 1, plan: existingPlan },
+      },
+      channelId: "telegram",
+    });
+
+    expect(removed?.channels).toEqual([]);
+
+    const rebuilt = await planner().buildRebuildPlanFromSandboxEntry({
+      sandboxName: "demo",
+      agent: "openclaw",
+      sandboxEntry: {
+        name: "demo",
+        messaging: { schemaVersion: 1, plan: removed! },
+      },
+      supportedChannelIds: ["telegram"],
+    });
+
+    expect(rebuilt).toMatchObject({
+      workflow: "rebuild",
+      channels: [],
+      disabledChannels: [],
+      credentialBindings: [],
+      networkPolicy: { presets: [], entries: [] },
+    });
+  });
+
   it("rebuilds from stored plan input values when config env is unavailable", async () => {
     const existingPlan = await withEnv(
       {
diff --git a/src/lib/messaging/compiler/workflow-planner.ts b/src/lib/messaging/compiler/workflow-planner.ts
index 768f2cfcc22..144e206519f 100644
--- a/src/lib/messaging/compiler/workflow-planner.ts
+++ b/src/lib/messaging/compiler/workflow-planner.ts
@@ -2,8 +2,6 @@
 // SPDX-License-Identifier: Apache-2.0
 
 import { MessagingHookRegistry } from "../hooks";
-import { hydrateDerivedSandboxMessagingPlanFields } from "../persistence";
-import { parseSandboxMessagingPlan } from "../plan-validation";
 import type {
   ChannelManifestRegistry,
   MessagingAgentId,
@@ -13,6 +11,8 @@ import type {
   SandboxMessagingPlan,
   SandboxMessagingRuntimeSetupPlan,
 } from "../manifest";
+import { hydrateDerivedSandboxMessagingPlanFields } from "../persistence";
+import { parseSandboxMessagingPlan } from "../plan-validation";
 import { planHostForward } from "./engines/host-forward-engine";
 import { planRuntimeSetup } from "./engines/runtime-setup-engine";
 import type { RenderTemplateReferenceResolver } from "./engines/template";
@@ -120,7 +120,7 @@ export class MessagingWorkflowPlanner {
     if (!existingPlan) return null;
 
     const filteredPlan = this.filterPlanChannelsToSupportedAllowlist(existingPlan, context);
-    if (!filteredPlan || filteredPlan.channels.length === 0) return null;
+    if (existingPlan.channels.length > 0 && filteredPlan.channels.length === 0) return null;
 
     return refreshDerivedPlanFields(
       setPlanDisabledChannels(
@@ -136,7 +136,7 @@ export class MessagingWorkflowPlanner {
   private filterPlanChannelsToSupportedAllowlist(
     plan: SandboxMessagingPlan,
     context: Pick<MessagingWorkflowPlannerBuildContext, "agent" | "supportedChannelIds">,
-  ): SandboxMessagingPlan | null {
+  ): SandboxMessagingPlan {
     if (!Array.isArray(context.supportedChannelIds)) return plan;
     const allowlist = new Set(context.supportedChannelIds);
     let filtered = plan;
diff --git a/test/e2e-scenario/live/agent-turn-latency.test.ts b/test/e2e-scenario/live/agent-turn-latency.test.ts
index 53e30cbdc36..d7ab7bb1d1f 100644
--- a/test/e2e-scenario/live/agent-turn-latency.test.ts
+++ b/test/e2e-scenario/live/agent-turn-latency.test.ts
@@ -5,6 +5,7 @@
 
 import fs from "node:fs";
 
+import { containsInteger42Answer } from "../../helpers/e2e-answer-assertions.ts";
 import { buildAvailabilityProbeEnv } from "../fixtures/availability-env.ts";
 import { resultText } from "../fixtures/clients/index.ts";
 import { trustedSandboxShellScript } from "../fixtures/clients/sandbox.ts";
@@ -88,9 +89,10 @@ test.skipIf(!shouldRunLiveE2EScenarios())(
     const openclaw = await openclawTurn(sandbox, apiKey);
     expect(openclaw.result.exitCode, resultText(openclaw.result)).toBe(0);
     assertNoOpenClawTransportErrors(resultText(openclaw.result));
-    expect(extractOpenClawAgentText(openclaw.result.stdout), resultText(openclaw.result)).toMatch(
-      /(^|[^0-9])42([^0-9]|$)/,
-    );
+    expect(
+      containsInteger42Answer(extractOpenClawAgentText(openclaw.result.stdout)),
+      resultText(openclaw.result),
+    ).toBe(true);
     expect(openclaw.elapsedMs).toBeLessThanOrEqual(MAX_TURN_SECONDS * 1000);
     results.openclaw = { elapsedMs: openclaw.elapsedMs };
 
@@ -152,7 +154,9 @@ test.skipIf(!shouldRunLiveE2EScenarios())(
     expect(hermesTurn.exitCode, resultText(hermesTurn)).toBe(0);
     const hermesResponse = responseBodyAndStatus(hermesTurn.stdout);
     expect(hermesResponse.status, resultText(hermesTurn)).toBe("200");
-    expect(chatContent(hermesResponse.body)).toMatch(/(^|[^0-9])42([^0-9]|$)/);
+    expect(containsInteger42Answer(chatContent(hermesResponse.body)), resultText(hermesTurn)).toBe(
+      true,
+    );
     expect(hermesMs).toBeLessThanOrEqual(MAX_TURN_SECONDS * 1000);
     results.hermes = { elapsedMs: hermesMs };
     await artifacts.writeJson("turn-latency-results.json", results);
diff --git a/test/e2e-scenario/live/full-e2e.test.ts b/test/e2e-scenario/live/full-e2e.test.ts
index 6eefd0d818d..778124d2d2e 100644
--- a/test/e2e-scenario/live/full-e2e.test.ts
+++ b/test/e2e-scenario/live/full-e2e.test.ts
@@ -7,6 +7,7 @@ import fs from "node:fs";
 import os from "node:os";
 import path from "node:path";
 
+import { containsInteger42Answer } from "../../helpers/e2e-answer-assertions.ts";
 import { buildAvailabilityProbeEnv } from "../fixtures/availability-env.ts";
 import { type HostCliClient } from "../fixtures/clients/host.ts";
 import { type SandboxClient, validateSandboxName } from "../fixtures/clients/sandbox.ts";
@@ -204,7 +205,9 @@ liveTest(
       },
     );
     expect(sandboxInference.exitCode, resultText(sandboxInference)).toBe(0);
-    expect(sandboxInference.stdout).toMatch(/(^|[^0-9])42([^0-9]|$)/);
+    expect(containsInteger42Answer(sandboxInference.stdout), resultText(sandboxInference)).toBe(
+      true,
+    );
 
     const logs = await repoNemoclaw(
       host,
diff --git a/test/e2e-scenario/live/gpu-double-onboard.test.ts b/test/e2e-scenario/live/gpu-double-onboard.test.ts
index e6091ddfc85..ace78c312dd 100644
--- a/test/e2e-scenario/live/gpu-double-onboard.test.ts
+++ b/test/e2e-scenario/live/gpu-double-onboard.test.ts
@@ -7,6 +7,7 @@ import fs from "node:fs";
 import os from "node:os";
 import path from "node:path";
 
+import { containsInteger42Answer } from "../../helpers/e2e-answer-assertions.ts";
 import { buildAvailabilityProbeEnv } from "../fixtures/availability-env.ts";
 import { type HostCliClient } from "../fixtures/clients/host.ts";
 import { type SandboxClient, validateSandboxName } from "../fixtures/clients/sandbox.ts";
@@ -145,7 +146,7 @@ async function expectSandboxInference42(
     },
   );
   expect(response.exitCode, resultText(response)).toBe(0);
-  expect(response.stdout).toMatch(/(^|[^0-9])42([^0-9]|$)/);
+  expect(containsInteger42Answer(response.stdout), resultText(response)).toBe(true);
 }
 
 liveTest(
diff --git a/test/e2e-scenario/live/issue-4462-scope-upgrade-approval.test.ts b/test/e2e-scenario/live/issue-4462-scope-upgrade-approval.test.ts
index 691017a74dc..7e40824968f 100644
--- a/test/e2e-scenario/live/issue-4462-scope-upgrade-approval.test.ts
+++ b/test/e2e-scenario/live/issue-4462-scope-upgrade-approval.test.ts
@@ -119,6 +119,13 @@ raise SystemExit(1)
 PY
 }
 
+contains_integer_42() {
+  local raw compact
+  raw="$(cat)"
+  compact="$(printf '%s' "$raw" | tr -d '[:space:]')"
+  grep -Eq '(^|[^0-9])42([^0-9]|$)' <<<"$compact"
+}
+
 assert_agent_scopes_without_admin() {
 python3 - <<'PY'
 import json, sys
@@ -159,7 +166,7 @@ if [ -z "$request_id" ]; then
     if printf '%s' "$state" | assert_agent_scopes_without_admin >/tmp/issue4462-approved-device.txt 2>/tmp/issue4462-approved-device.err; then
       echo "SCOPE_ALREADY_APPROVED=$(cat /tmp/issue4462-approved-device.txt)"
     elif [ "$trigger_rc" -eq 0 ] && ! grep -Eiq 'EMBEDDED FALLBACK|scope upgrade pending approval|pairing required|fallbackFrom[": ]+gateway|transport[": ]+embedded' /tmp/issue4462-trigger-agent.log \
-      && grep -Eq '(^|[^0-9])42([^0-9]|$)' /tmp/issue4462-trigger-agent.log; then
+      && contains_integer_42 </tmp/issue4462-trigger-agent.log; then
       echo "TRIGGER_COMPLETED_WITHOUT_PENDING_SCOPE_UPGRADE"
       echo "ISSUE_4462_SCOPE_UPGRADE_OK device=trigger-completed request=not-reproduced"
       exit 0
@@ -209,7 +216,7 @@ if grep -Eiq 'EMBEDDED FALLBACK|scope upgrade pending approval|pairing required|
   cat /tmp/issue4462-final-agent.log >&2
   exit 7
 fi
-if ! grep -Eq '(^|[^0-9])42([^0-9]|$)' /tmp/issue4462-final-agent.log; then
+if ! contains_integer_42 </tmp/issue4462-final-agent.log; then
   echo "FINAL_AGENT_MISSING_42" >&2
   cat /tmp/issue4462-final-agent.log >&2
   exit 8
diff --git a/test/e2e-scenario/live/launchable-smoke.test.ts b/test/e2e-scenario/live/launchable-smoke.test.ts
index b9660950941..e81a30d36ca 100644
--- a/test/e2e-scenario/live/launchable-smoke.test.ts
+++ b/test/e2e-scenario/live/launchable-smoke.test.ts
@@ -6,8 +6,9 @@ import fs from "node:fs";
 import os from "node:os";
 import path from "node:path";
 
-import { buildAvailabilityProbeEnv } from "../fixtures/availability-env.ts";
+import { containsInteger42Answer } from "../../helpers/e2e-answer-assertions.ts";
 import type { ArtifactSink } from "../fixtures/artifacts.ts";
+import { buildAvailabilityProbeEnv } from "../fixtures/availability-env.ts";
 import type { HostCliClient } from "../fixtures/clients/host.ts";
 import { validateSandboxName } from "../fixtures/clients/sandbox.ts";
 import { expect, test } from "../fixtures/e2e-test.ts";
@@ -463,7 +464,7 @@ runLaunchableSmokeTest(
     ).toBe(0);
     const agentReply = parseAgentText(agent.stdout);
     expect(
-      /(^|[^0-9])42([^0-9]|$)/.test(agentReply),
+      containsInteger42Answer(agentReply),
       `expected agent reply to contain 42; rc=${agent.exitCode}; reply='${agentReply.slice(0, 200)}'; stdout='${agent.stdout.slice(0, 300)}'; stderr='${agent.stderr.slice(0, 300)}'`,
     ).toBe(true);
 
diff --git a/test/e2e-scenario/live/openclaw-tui-chat-correlation.test.ts b/test/e2e-scenario/live/openclaw-tui-chat-correlation.test.ts
index 9759f12c261..26034af662a 100644
--- a/test/e2e-scenario/live/openclaw-tui-chat-correlation.test.ts
+++ b/test/e2e-scenario/live/openclaw-tui-chat-correlation.test.ts
@@ -19,6 +19,7 @@ import { mkdtempSync, rmSync, writeFileSync } from "node:fs";
 import { tmpdir } from "node:os";
 import { join } from "node:path";
 
+import { containsReplyTokenAllowingWhitespace } from "../../helpers/e2e-answer-assertions.ts";
 import { buildAvailabilityProbeEnv } from "../fixtures/availability-env.ts";
 import { type SandboxClient, trustedSandboxShellScript } from "../fixtures/clients/sandbox.ts";
 import { expect, test } from "../fixtures/e2e-test.ts";
@@ -160,7 +161,7 @@ function analyzeIssue2603Trace({
   const finalReplyCounts = new Map<string, number>();
   for (const [replyToken, expectedRunId] of expectedRunByReplyToken) {
     for (const event of chatEvents) {
-      if (!event.text.includes(replyToken)) continue;
+      if (!containsReplyTokenAllowingWhitespace(event.text, replyToken)) continue;
       visibleReplyCounts.set(replyToken, (visibleReplyCounts.get(replyToken) ?? 0) + 1);
       if (event.state === "final") {
         finalReplyCounts.set(replyToken, (finalReplyCounts.get(replyToken) ?? 0) + 1);
@@ -188,7 +189,7 @@ function analyzeIssue2603Trace({
     .filter((event) => event.state === "final")
     .flatMap((event) =>
       sentRuns
-        .filter((entry) => event.text.includes(entry.replyToken))
+        .filter((entry) => containsReplyTokenAllowingWhitespace(event.text, entry.replyToken))
         .map((entry) => entry.replyToken),
     );
 
@@ -288,8 +289,12 @@ function textFromMessage(message) {
   return content.map((part) => part && typeof part === "object" && typeof part.text === "string" ? part.text : "").filter(Boolean).join("\n");
 }
 
+function compactReplyTokenText(value) {
+  return String(value || "").replace(/\s+/g, "");
+}
+
 function sawAllReplies(replyTokens) {
-  return replyTokens.every((token) => events.some((event) => event.event === "chat" && textFromMessage(event.payload?.message).includes(token)));
+  return replyTokens.every((token) => events.some((event) => event.event === "chat" && compactReplyTokenText(textFromMessage(event.payload?.message)).includes(compactReplyTokenText(token))));
 }
 
 ws.on("message", (data) => {
diff --git a/test/e2e-scenario/live/sandbox-operations.test.ts b/test/e2e-scenario/live/sandbox-operations.test.ts
index acaf07de8f5..13a7e1edd5e 100644
--- a/test/e2e-scenario/live/sandbox-operations.test.ts
+++ b/test/e2e-scenario/live/sandbox-operations.test.ts
@@ -13,6 +13,7 @@ import fs from "node:fs";
 import os from "node:os";
 import path from "node:path";
 
+import { containsInteger42Answer } from "../../helpers/e2e-answer-assertions.ts";
 import { buildAvailabilityProbeEnv } from "../fixtures/availability-env.ts";
 import type { HostCliClient } from "../fixtures/clients/host.ts";
 import { type SandboxClient, trustedSandboxShellScript } from "../fixtures/clients/sandbox.ts";
@@ -266,7 +267,7 @@ async function assertAgentCanAnswer(host: HostCliClient, sandboxName: string): P
   );
   const reply = parseOpenClawAgentText(result.stdout);
   expectExitZero(result, "openclaw agent --json");
-  expect(reply, resultText(result)).toMatch(/(^|[^0-9])42([^0-9]|$)/);
+  expect(containsInteger42Answer(reply), resultText(result)).toBe(true);
 }
 
 async function assertStatusFields(host: HostCliClient, sandboxName: string): Promise<void> {
diff --git a/test/e2e/lib/openclaw-json.sh b/test/e2e/lib/openclaw-json.sh
index 8f17bab69fa..74bfd814b6a 100755
--- a/test/e2e/lib/openclaw-json.sh
+++ b/test/e2e/lib/openclaw-json.sh
@@ -8,6 +8,12 @@
 # exact envelope shape. This also tolerates wrapper output before the JSON blob
 # but intentionally ignores metadata fields so IDs, durations, session names,
 # and model/provider details cannot satisfy reply assertions.
+e2e_text_contains_integer_42() {
+  local compact
+  compact="$(printf '%s' "${1:-}" | tr -d '[:space:]')"
+  grep -qE '(^|[^0-9])42([^0-9]|$)' <<<"$compact"
+}
+
 parse_openclaw_agent_text() {
   python3 -c '
 import json
diff --git a/test/e2e/test-agent-turn-latency-e2e.sh b/test/e2e/test-agent-turn-latency-e2e.sh
index 694317f7a1d..306a7830adf 100755
--- a/test/e2e/test-agent-turn-latency-e2e.sh
+++ b/test/e2e/test-agent-turn-latency-e2e.sh
@@ -412,7 +412,7 @@ run_openclaw_turn() {
     return
   fi
 
-  if grep -qE '(^|[^0-9])42([^0-9]|$)' <<<"$reply"; then
+  if e2e_text_contains_integer_42 "$reply"; then
     pass "OpenClaw: real agent turn returned 42 in $(duration_s "$OPENCLAW_TURN_MS")"
     assert_latency_under_cap "OpenClaw" "$OPENCLAW_TURN_MS"
   else
@@ -463,7 +463,7 @@ print(json.dumps({
     return
   fi
 
-  if grep -qE '(^|[^0-9])42([^0-9]|$)' <<<"$content"; then
+  if e2e_text_contains_integer_42 "$content"; then
     pass "Hermes: real daemon turn returned 42 in $(duration_s "$HERMES_TURN_MS")"
     assert_latency_under_cap "Hermes" "$HERMES_TURN_MS"
   else
diff --git a/test/e2e/test-full-e2e.sh b/test/e2e/test-full-e2e.sh
index 6d144f13ae3..9a2aa63953a 100755
--- a/test/e2e/test-full-e2e.sh
+++ b/test/e2e/test-full-e2e.sh
@@ -447,7 +447,7 @@ rm -f "$ssh_config"
 
 agent_reply=$(printf '%s' "$agent_response" | parse_openclaw_agent_text 2>/dev/null) || true
 
-if grep -qE "(^|[^0-9])42([^0-9]|$)" <<<"$agent_reply"; then
+if e2e_text_contains_integer_42 "$agent_reply"; then
   pass "[LIVE] openclaw agent: model answered 6×7=42 through openclaw → inference.local"
 else
   fail "[LIVE] openclaw agent: expected '42' in agent reply, got: ${agent_reply:0:200}"
diff --git a/test/e2e/test-issue-4462-scope-upgrade-approval.sh b/test/e2e/test-issue-4462-scope-upgrade-approval.sh
index a1fc4f34416..287e1b7eb1a 100755
--- a/test/e2e/test-issue-4462-scope-upgrade-approval.sh
+++ b/test/e2e/test-issue-4462-scope-upgrade-approval.sh
@@ -1074,7 +1074,7 @@ openclaw agent --agent main --json --session-id "$session_id" \
     last_agent_detail="agent exited ${final_rc}: ${final_output:0:500}"
   elif ! grep -q '^__URL_FOR_FINAL_AGENT__=ws://' <<<"$final_output"; then
     last_agent_detail="agent command did not preserve OPENCLAW_GATEWAY_URL: ${final_output:0:500}"
-  elif grep -qE '(^|[^0-9])42([^0-9]|$)' <<<"$reply"; then
+  elif e2e_text_contains_integer_42 "$reply"; then
     agent_ok=1
     pass "approved openclaw agent turn answered through gateway mode"
     break
diff --git a/test/e2e/test-launchable-smoke.sh b/test/e2e/test-launchable-smoke.sh
index 73d993f227d..dcddf40b73e 100755
--- a/test/e2e/test-launchable-smoke.sh
+++ b/test/e2e/test-launchable-smoke.sh
@@ -539,7 +539,7 @@ rm -f "$ssh_config" "$agent_stderr_file"
 
 agent_reply=$(printf '%s' "$agent_response" | parse_openclaw_agent_text 2>/dev/null) || true
 
-if grep -qE "(^|[^0-9])42([^0-9]|$)" <<<"$agent_reply"; then
+if e2e_text_contains_integer_42 "$agent_reply"; then
   pass "[LIVE] openclaw agent: model answered 6×7=42 through openclaw → inference.local"
 else
   fail "[LIVE] openclaw agent: expected '42' in agent reply; rc=${agent_rc}; reply='${agent_reply:0:200}'; stdout='${agent_response:0:300}'; stderr='${agent_stderr:0:300}'"
diff --git a/test/e2e/test-sandbox-operations.sh b/test/e2e/test-sandbox-operations.sh
index 98d1b015379..5be9c99d277 100755
--- a/test/e2e/test-sandbox-operations.sh
+++ b/test/e2e/test-sandbox-operations.sh
@@ -387,7 +387,7 @@ test_sbx_02_connect_chat() {
   local reply
   reply=$(printf '%s' "$raw" | parse_openclaw_agent_text 2>/dev/null) || true
 
-  if [[ $rc -eq 0 && -n "$reply" ]] && echo "$reply" | grep -qE "(^|[^0-9])42([^0-9]|$)"; then
+  if [[ $rc -eq 0 && -n "$reply" ]] && e2e_text_contains_integer_42 "$reply"; then
     pass "TC-SBX-02: Agent computed 6×7=42 through openclaw → inference.local"
   else
     fail "TC-SBX-02: Connect & Chat" "Expected '42' in agent reply (rc=$rc); reply='${reply:0:200}'; raw output='${raw:0:200}'"
diff --git a/test/helpers/e2e-answer-assertions.test.ts b/test/helpers/e2e-answer-assertions.test.ts
new file mode 100644
index 00000000000..4e20cb02080
--- /dev/null
+++ b/test/helpers/e2e-answer-assertions.test.ts
@@ -0,0 +1,28 @@
+// SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+import { describe, expect, it } from "vitest";
+
+import {
+  compactAnswerText,
+  containsInteger42Answer,
+  containsReplyTokenAllowingWhitespace,
+} from "./e2e-answer-assertions.ts";
+
+describe("E2E answer assertions", () => {
+  it("normalizes harmless model-inserted whitespace", () => {
+    expect(compactAnswerText("4\n2")).toBe("42");
+    expect(containsInteger42Answer("4\n2")).toBe(true);
+    expect(containsInteger42Answer("The answer is 4\n2.")).toBe(true);
+  });
+
+  it("does not match unrelated integers after whitespace normalization", () => {
+    expect(containsInteger42Answer("142")).toBe(false);
+    expect(containsInteger42Answer("420")).toBe(false);
+  });
+
+  it("matches deterministic reply tokens split by streaming whitespace", () => {
+    expect(containsReplyTokenAllowingWhitespace("A\n2603-REPLY", "A2603-REPLY")).toBe(true);
+    expect(containsReplyTokenAllowingWhitespace("B 2603-REPLY", "B2603-REPLY")).toBe(true);
+  });
+});
diff --git a/test/helpers/e2e-answer-assertions.ts b/test/helpers/e2e-answer-assertions.ts
new file mode 100644
index 00000000000..b72176d52b5
--- /dev/null
+++ b/test/helpers/e2e-answer-assertions.ts
@@ -0,0 +1,14 @@
+// SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+export function compactAnswerText(text: string): string {
+  return text.replace(/\s+/g, "");
+}
+
+export function containsInteger42Answer(text: string): boolean {
+  return /(^|[^0-9])42([^0-9]|$)/.test(compactAnswerText(text));
+}
+
+export function containsReplyTokenAllowingWhitespace(text: string, replyToken: string): boolean {
+  return compactAnswerText(text).includes(compactAnswerText(replyToken));
+}
diff --git a/test/openclaw-tui-chat-correlation.test.ts b/test/openclaw-tui-chat-correlation.test.ts
index dc006296886..aca870d36d1 100644
--- a/test/openclaw-tui-chat-correlation.test.ts
+++ b/test/openclaw-tui-chat-correlation.test.ts
@@ -1,14 +1,16 @@
 // SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 // SPDX-License-Identifier: Apache-2.0
 
-import { execFileSync } from "node:child_process";
 import type { ExecFileSyncOptionsWithStringEncoding } from "node:child_process";
+import { execFileSync } from "node:child_process";
 import { randomUUID } from "node:crypto";
 import fs from "node:fs";
 import os from "node:os";
 import path from "node:path";
 import { describe, expect, it } from "vitest";
 
+import { containsReplyTokenAllowingWhitespace } from "./helpers/e2e-answer-assertions.ts";
+
 const LIVE_REPRO_ENV = "NEMOCLAW_ISSUE_2603_LIVE";
 const LIVE_SANDBOX_ENV = "NEMOCLAW_ISSUE_2603_SANDBOX";
 const LIVE_SCRIPT_NAME = "openclaw-issue2603-chat-correlation.cjs";
@@ -150,7 +152,7 @@ function analyzeIssue2603Trace({
   const finalReplyCounts = new Map<string, number>();
   for (const [replyToken, expectedRunId] of expectedRunByReplyToken) {
     for (const event of chatEvents) {
-      if (!event.text.includes(replyToken)) continue;
+      if (!containsReplyTokenAllowingWhitespace(event.text, replyToken)) continue;
       visibleReplyCounts.set(replyToken, (visibleReplyCounts.get(replyToken) ?? 0) + 1);
       if (event.state === "final") {
         finalReplyCounts.set(replyToken, (finalReplyCounts.get(replyToken) ?? 0) + 1);
@@ -423,8 +425,12 @@ function textFromMessage(message) {
   return content.map((part) => part && typeof part === "object" && typeof part.text === "string" ? part.text : "").filter(Boolean).join("\n");
 }
 
+function compactReplyTokenText(value) {
+  return String(value || "").replace(/\s+/g, "");
+}
+
 function sawAllReplies(replyTokens) {
-  return replyTokens.every((token) => events.some((event) => event.event === "chat" && textFromMessage(event.payload?.message).includes(token)));
+  return replyTokens.every((token) => events.some((event) => event.event === "chat" && compactReplyTokenText(textFromMessage(event.payload?.message)).includes(compactReplyTokenText(token))));
 }
 
 ws.on("message", (data) => {
@@ -611,6 +617,38 @@ describe("OpenClaw TUI chat correlation regression (#2603)", () => {
     ]);
   });
 
+  it("matches deterministic reply tokens split by harmless model whitespace", () => {
+    const analysis = analyzeIssue2603Trace({
+      sentRuns: [
+        {
+          promptToken: "A2603",
+          replyToken: "A2603-REPLY",
+          runId: "split-reply-run",
+          message: "A2603: Reply exactly A2603-REPLY and nothing else.",
+        },
+      ],
+      events: [
+        {
+          event: "chat",
+          payload: {
+            runId: "split-reply-run",
+            state: "final",
+            message: { role: "assistant", content: [{ type: "text", text: "A\n2603-REPLY" }] },
+          },
+        },
+      ],
+      historyMessages: [
+        {
+          role: "user",
+          content: [{ type: "text", text: "A2603: Reply exactly A2603-REPLY and nothing else." }],
+        },
+      ],
+    });
+
+    expect(analysis.missingReplies).toEqual([]);
+    expect(analysis.uncorrelatedReplies).toEqual([]);
+  });
+
   it("does not classify a streamed delta plus final for the same run as a duplicate reply", () => {
     const analysis = analyzeIssue2603Trace({
       sentRuns: [