fix(opencode): never expose a port we don't manage to the process killer

The Electron-side OpenCode killer kills by port (lsof + kill -KILL).
getOpenCodeProcessInfo returned openCodePort unconditionally, so for an
external/attached OpenCode (e.g. a user's own server on 4096) the only thing
stopping the killer from taking it down was the separate `managed` flag — a
single weak signal guarding a destructive action.

Withhold pid/port unless we actually manage the process, so the killer has no
target even if `managed` is ever miscomputed. Managed flow is unchanged.

packages/web/server/index.js

@@ -1327,11 +1327,20 @@ async function main(options = {}) {
     }),
     isReady: () => isOpenCodeReady,
     restartOpenCode: () => restartOpenCode(),
-    getOpenCodeProcessInfo: () => ({
+    getOpenCodeProcessInfo: () => {
-      managed: Boolean((openCodeProcess || openCodePort) && !ENV_SKIP_OPENCODE_START && !isExternalOpenCode),
+      const managed = Boolean((openCodeProcess || openCodePort) && !ENV_SKIP_OPENCODE_START && !isExternalOpenCode);
-      pid: typeof openCodeProcess?.pid === 'number' ? openCodeProcess.pid : null,
+      // Only ever expose pid/port for a server WE manage. The Electron-side
-      port: openCodePort,
+      // killer kills by port (lsof + kill -KILL), so returning a port we don't
-    }),
+      // own — e.g. an external/desktop OpenCode on 4096 we attached to — would
+      // let a single miscomputed `managed` flag take down the user's separate
+      // server. Structurally withhold what isn't ours so the killer has no
+      // target, instead of relying on the flag check alone.
+      return {
+        managed,
+        pid: managed && typeof openCodeProcess?.pid === 'number' ? openCodeProcess.pid : null,
+        port: managed ? openCodePort : null,
+      };
+    },
     stop: (shutdownOptions = {}) =>
       gracefulShutdown({ exitProcess: shutdownOptions.exitProcess ?? false })
   };