diff --git a/src/adapters/cursor/tool-result-normalize.ts b/src/adapters/cursor/tool-result-normalize.ts
index c89e247e9a..ab5b567b1d 100644
--- a/src/adapters/cursor/tool-result-normalize.ts
+++ b/src/adapters/cursor/tool-result-normalize.ts
@@ -35,23 +35,26 @@ function isNodeReplOrComputerUseTool(toolName?: string, toolNamespace?: string):
  * reads the blank [tool_result], concludes prior results were lost, and spirals into
  * re-orientation retries (devlog 260826_cursor_responses_gap, live subagent transcripts).
  */
+const CODEX_EXEC_BRIDGE_TOOL_NAMES = new Set([
+  "exec",
+  "exec_command",
+  "shell_command",
+  "shell",
+  "local_shell",
+  "container.exec",
+]);
+
 function isCodexExecBridgeTool(toolName?: string, toolNamespace?: string): boolean {
-  if (toolNamespace && toolNamespace.includes("opencodex-responses")) return true;
   if (!toolName) return false;
   const lower = toolName.toLowerCase();
-  return (
-    lower === "exec"
-    || lower === "exec_command"
-    || lower === "shell_command"
-    // Codex CLI/desktop native tool names: the multi-round "이전 출력이 비어 있어 처음부터"
-    // restart loop reproduced via codex exec because `shell` was not in this set
-    // (devlog 260826 gap-8 QA round 2).
-    || lower === "shell"
-    || lower === "local_shell"
-    || lower === "container.exec"
-    || lower.startsWith("mcp_opencodex-responses_")
-    || lower.startsWith("mcp__opencodex-responses__")
-  );
+  if (toolNamespace) {
+    return toolNamespace === "opencodex-responses" && CODEX_EXEC_BRIDGE_TOOL_NAMES.has(lower);
+  }
+  if (CODEX_EXEC_BRIDGE_TOOL_NAMES.has(lower)) return true;
+  return [...CODEX_EXEC_BRIDGE_TOOL_NAMES].some(name => (
+    lower === `mcp_opencodex-responses_${name}`
+    || lower === `mcp__opencodex-responses__${name}`
+  ));
 }
 
 /** Failure states the Computer Use / node_repl runtime reports as PLAIN TEXT inside a non-error result. */
@@ -108,7 +111,7 @@ export function normalizeCursorToolResultText(
   if (isCodexExecBridgeTool(options.toolName, options.toolNamespace) && EMPTY_EXEC_OUTPUT_REGEX.test(text.trim())) {
     return {
       text: "[empty output: the exec cell completed but emitted nothing. This is NOT lost context and NOT a blocked tool — in code mode call text(...) or notify(...) on any value you need to see (a bare await tools.exec_command(...) is not echoed automatically); in shell mode the command simply printed nothing. Do not re-run the same call expecting different output.]",
-      isError: false,
+      isError,
       changed: true,
     };
   }
diff --git a/tests/cursor-exec-empty-result.test.ts b/tests/cursor-exec-empty-result.test.ts
index 096aa01f9c..81eb7dd433 100644
--- a/tests/cursor-exec-empty-result.test.ts
+++ b/tests/cursor-exec-empty-result.test.ts
@@ -16,6 +16,40 @@ describe("codex exec bridge empty-result normalization (devlog 260826 gap-7)", (
     expect(out.text).toContain("empty output");
   });
 
+  test("reserved namespace requires an exact exec bridge name", () => {
+    const out = normalizeCursorToolResultText("", {
+      toolName: "read_file",
+      toolNamespace: "opencodex-responses",
+    });
+    expect(out).toEqual({ text: "", isError: false, changed: false });
+  });
+
+  test("exec names in unrelated namespaces stay byte-identical", () => {
+    const out = normalizeCursorToolResultText("", {
+      toolName: "exec",
+      toolNamespace: "thirdparty",
+      isError: true,
+    });
+    expect(out).toEqual({ text: "", isError: true, changed: false });
+  });
+
+  test("namespace and display aliases must match the reserved identity exactly", () => {
+    for (const options of [
+      { toolName: "read_file", toolNamespace: "evil-opencodex-responses-suffix" },
+      { toolName: "mcp_opencodex-responses_read_file" },
+      { toolName: "mcp__opencodex-responses__read_file" },
+    ]) {
+      expect(normalizeCursorToolResultText("", options).changed).toBe(false);
+    }
+  });
+
+  test("existing exec errors remain errors after empty-output guidance is added", () => {
+    const out = normalizeCursorToolResultText("", { toolName: "exec", isError: true });
+    expect(out.changed).toBe(true);
+    expect(out.isError).toBe(true);
+    expect(out.text).toContain("empty output");
+  });
+
   test("shell_command empty output routes too", () => {
     const out = normalizeCursorToolResultText("<empty>", { toolName: "shell_command" });
     expect(out.changed).toBe(true);