wtagent 0.4.0 → 1.0.0

package/src/browser/grok-web-adapter.js

@@ -1,5 +1,6 @@
 import { BaseWebAdapter, firstVisible } from "./base-web-adapter.js";
 import { isUsageLimitNotice } from "../shared/usage-limit.js";
+import { readRenderedBlocks } from "./rendered-text.js";
 
 export { isConnectionLostError } from "./base-web-adapter.js";
 
@@ -20,7 +21,11 @@ export class GrokWebAdapter extends BaseWebAdapter {
 
   composerLocators() {
     return [
-      this.page.locator("main textarea").first(),
+      // Grok replaces its initial textarea with a role=textbox editor during
+      // hydration. Keep both variants in one live locator so a composer found
+      // just before that replacement does not become an empty stale selector
+      // by the time sendMessage fills it.
+      this.page.locator('main textarea, main [role="textbox"]').first(),
       this.page.locator("textarea[aria-label]").first(),
       this.page.locator('[role="textbox"]').first(),
     ];
@@ -84,13 +89,40 @@ export class GrokWebAdapter extends BaseWebAdapter {
     );
   }
 
+  async userMessageText(message) {
+    const markdown = message.locator(".response-content-markdown");
+    if (await markdown.count().catch(() => 0) > 0) {
+      const exact = await markdown.evaluateAll(readRenderedBlocks);
+      if (exact.trim()) return exact;
+    }
+    return await super.userMessageText(message);
+  }
+
   async messageIdentity(message) {
     const testId = await message.getAttribute("data-testid").catch(() => null);
     const aria = await message.getAttribute("aria-label").catch(() => null);
     const scope = testId === "user-message" || aria === "You"
       ? this.userMessages()
       : this.assistantMessages();
-    return { id: null, turn: await scope.count().catch(() => 0) };
+    // Grok wraps every visible user/assistant bubble in a stable response UUID:
+    //
+    //   <div id="response-..." data-scroll-anchor-root>
+    //     <div data-testid="user-message|assistant-message">...</div>
+    //   </div>
+    //
+    // Keep the count fallback for older DOM variants, but prefer the wrapper
+    // identity whenever it is present. Besides making normal multi-turn
+    // correlation stronger, this lets a fresh chat safely follow Grok's late
+    // root -> /c/... navigation after the user bubble has already committed.
+    const id = await message.evaluate((element) => (
+      element.closest('[data-scroll-anchor-root][id^="response-"]')?.id
+        ?? element.closest('[id^="response-"]')?.id
+        ?? null
+    )).catch(() => null);
+    return {
+      id: id || null,
+      turn: await scope.count().catch(() => 0),
+    };
   }
 
   isNewAssistantIdentity({ turn, text }) {
@@ -104,7 +136,7 @@ export class GrokWebAdapter extends BaseWebAdapter {
   async assistantText(message) {
     const response = message.locator(".response-content-markdown");
     if (await response.count().catch(() => 0) > 0) {
-      return await response.last().innerText().catch(() => "");
+      return await response.evaluateAll(readRenderedBlocks);
     }
     return await message.innerText().catch(() => "");
   }

package/src/browser/rendered-text.js

@@ -0,0 +1,11 @@
+// Passed directly to Locator.evaluateAll: keep this function self-contained.
+// Read connected nodes so innerText keeps layout whitespace and omits hidden
+// descendants. Cloning first changes innerText into textContent-like output.
+export function readRenderedBlocks(elements) {
+  const roots = elements.filter((element) => (
+    !elements.some((other) => other !== element && other.contains(element))
+  ));
+  return roots.filter((element) => element.getClientRects().length > 0)
+    .map((element) => element.innerText)
+    .join("\n");
+}

package/src/cli/i18n.js

@@ -20,6 +20,7 @@ const MESSAGES = Object.freeze({
     "option.project": "Project directory",
     "option.model": "Web AI provider: {providers} (default: {defaultProvider})",
     "option.once": "Run a single request and exit instead of a conversation",
+    "option.type": "Request type: agent (default) or music (Gemini only)",
     "option.timeout": "Maximum wait for one model response (default: 10 minutes)",
     "option.noMinimize": "Keep the Chrome window visible instead of minimizing it",
     "option.debug": "Write browser diagnostics",
@@ -42,7 +43,9 @@ const MESSAGES = Object.freeze({
     "login.savingProfile": "Closing native Chrome and saving the dedicated profile...",
     "login.verified": "{provider} login verified through a fresh CDP connection.",
     "login.stillGuest": "{provider} is still in guest mode. Reopening native Chrome; complete the final {provider} sign-in/continue step.",
-    "model.chooseInBrowser": "{provider} is signed in. Choose the model you want to use in the browser if desired, then press Enter here. Press Enter immediately to keep the currently selected model.",
+    "model.chooseInBrowser": "{provider} is signed in.\nYou can change the model in the browser.\nPress Enter to continue, or wait 5 seconds.",
+    "model.continueCountdown": "Continuing in {seconds}s…",
+    "model.continuing": "Continuing…",
     "model.setupCancelled": "Model selection confirmation cancelled.",
     "logout.none": "No Chrome profile found at {profileDir}; already logged out.",
     "logout.refuse": "Refusing to delete {profileDir}: it does not look like a wtagent Chrome profile.",
@@ -106,6 +109,7 @@ const MESSAGES = Object.freeze({
     "option.project": "项目目录",
     "option.model": "网页 AI 提供商：{providers}（默认：{defaultProvider}）",
     "option.once": "只执行一次请求并退出，不进入连续对话",
+    "option.type": "请求类型：agent（默认）或 music（仅 Gemini）",
     "option.timeout": "单次模型响应最大等待时间（默认：10 分钟）",
     "option.noMinimize": "保持 Chrome 窗口可见，不自动最小化",
     "option.debug": "写入浏览器诊断信息",
@@ -128,7 +132,9 @@ const MESSAGES = Object.freeze({
     "login.savingProfile": "正在关闭原生 Chrome 并保存专用配置...",
     "login.verified": "已通过新的 CDP 连接验证 {provider} 登录状态。",
     "login.stillGuest": "{provider} 仍处于访客模式。正在重新打开 Chrome；请完成最后的登录/继续步骤。",
-    "model.chooseInBrowser": "{provider} 已登录。如需切换模型，请直接在浏览器网页中人工选择；完成后在此按 Enter。若保持当前已选模型，直接按 Enter 即可。",
+    "model.chooseInBrowser": "{provider} 已登录。\n如需切换模型，请在浏览器中选择。\n按 Enter 立即继续，或等待 5 秒自动继续。",
+    "model.continueCountdown": "{seconds} 秒后自动继续…",
+    "model.continuing": "正在继续…",
     "model.setupCancelled": "模型选择确认已取消。",
     "logout.none": "在 {profileDir} 未找到 Chrome 配置；当前已是登出状态。",
     "logout.refuse": "拒绝删除 {profileDir}：它看起来不是 wtagent 的 Chrome 配置目录。",
@@ -192,6 +198,7 @@ const MESSAGES = Object.freeze({
     "option.project": "專案目錄",
     "option.model": "網頁 AI 提供者：{providers}（預設：{defaultProvider}）",
     "option.once": "只執行一次請求並退出，不進入連續對話",
+    "option.type": "請求類型：agent（預設）或 music（僅 Gemini）",
     "option.timeout": "單次模型回應最大等待時間（預設：10 分鐘）",
     "option.noMinimize": "保持 Chrome 視窗可見，不自動最小化",
     "option.debug": "寫入瀏覽器診斷資訊",
@@ -209,7 +216,9 @@ const MESSAGES = Object.freeze({
     "login.savingProfile": "正在關閉原生 Chrome 並儲存專用設定檔...",
     "login.verified": "已透過新的 CDP 連線驗證 {provider} 登入狀態。",
     "login.stillGuest": "{provider} 仍處於訪客模式。正在重新開啟 Chrome；請完成最後的登入/繼續步驟。",
-    "model.chooseInBrowser": "{provider} 已登入。如需切換模型，請直接在瀏覽器頁面中自行選擇；完成後在此按 Enter。若保持目前已選模型，直接按 Enter 即可。",
+    "model.chooseInBrowser": "{provider} 已登入。\n如需切換模型，請在瀏覽器中選擇。\n按 Enter 立即繼續，或等待 5 秒自動繼續。",
+    "model.continueCountdown": "{seconds} 秒後自動繼續…",
+    "model.continuing": "正在繼續…",
     "model.setupCancelled": "模型選擇確認已取消。",
     "task.youPrompt": "你 ›",
     "option.yes": "略過確認提示",
@@ -278,6 +287,7 @@ const MESSAGES = Object.freeze({
     "option.project": "プロジェクトディレクトリ",
     "option.model": "Web AI プロバイダー: {providers}（既定: {defaultProvider}）",
     "option.once": "会話を継続せず、1 回のリクエストを実行して終了",
+    "option.type": "リクエストの種類: agent（既定）または music（Gemini のみ）",
     "option.timeout": "1 回のモデル応答の最大待機時間（既定: 10 分）",
     "option.noMinimize": "Chrome ウィンドウを最小化せず表示したままにする",
     "option.debug": "ブラウザー診断を書き出す",
@@ -295,7 +305,9 @@ const MESSAGES = Object.freeze({
     "login.savingProfile": "ネイティブ Chrome を閉じ、専用プロファイルを保存しています...",
     "login.verified": "新しい CDP 接続で {provider} のログインを確認しました。",
     "login.stillGuest": "{provider} はまだゲスト状態です。Chrome を再度開きますので、最後のサインイン/続行手順を完了してください。",
-    "model.chooseInBrowser": "{provider} にログイン済みです。必要ならブラウザー上で使用するモデルを手動で選択し、ここで Enter を押してください。現在のモデルを使う場合は、そのまま Enter を押します。",
+    "model.chooseInBrowser": "{provider} にログイン済みです。\n必要ならブラウザーでモデルを変更できます。\nEnter で続行、または 5 秒後に自動で続行します。",
+    "model.continueCountdown": "{seconds} 秒後に続行…",
+    "model.continuing": "続行中…",
     "model.setupCancelled": "モデル選択の確認をキャンセルしました。",
     "task.youPrompt": "あなた ›",
     "option.yes": "確認プロンプトを省略",
@@ -364,6 +376,7 @@ const MESSAGES = Object.freeze({
     "option.project": "프로젝트 디렉터리",
     "option.model": "웹 AI 제공자: {providers} (기본값: {defaultProvider})",
     "option.once": "대화를 계속하지 않고 요청 한 번만 실행한 뒤 종료",
+    "option.type": "요청 유형: agent(기본값) 또는 music(Gemini 전용)",
     "option.timeout": "모델 응답 1회의 최대 대기 시간 (기본값: 10분)",
     "option.noMinimize": "Chrome 창을 최소화하지 않고 표시",
     "option.debug": "브라우저 진단 정보 기록",
@@ -381,7 +394,9 @@ const MESSAGES = Object.freeze({
     "login.savingProfile": "네이티브 Chrome을 닫고 전용 프로필을 저장하는 중...",
     "login.verified": "새 CDP 연결을 통해 {provider} 로그인을 확인했습니다.",
     "login.stillGuest": "{provider}가 아직 게스트 모드입니다. Chrome을 다시 열고 마지막 로그인/계속 단계를 완료하세요.",
-    "model.chooseInBrowser": "{provider}에 로그인되었습니다. 원하는 경우 브라우저에서 사용할 모델을 직접 선택한 뒤 여기서 Enter를 누르세요. 현재 선택된 모델을 유지하려면 바로 Enter를 누르면 됩니다.",
+    "model.chooseInBrowser": "{provider}에 로그인되었습니다.\n필요하면 브라우저에서 모델을 변경하세요.\nEnter를 누르거나 5초 후 자동으로 계속합니다.",
+    "model.continueCountdown": "{seconds}초 후 계속…",
+    "model.continuing": "계속하는 중…",
     "model.setupCancelled": "모델 선택 확인이 취소되었습니다.",
     "task.youPrompt": "나 ›",
     "option.yes": "확인 프롬프트 건너뛰기",
@@ -450,6 +465,7 @@ const MESSAGES = Object.freeze({
     "option.project": "Directorio del proyecto",
     "option.model": "Proveedor de IA web: {providers} (predeterminado: {defaultProvider})",
     "option.once": "Ejecutar una sola solicitud y salir en lugar de mantener una conversación",
+    "option.type": "Tipo de solicitud: agent (predeterminado) o music (solo Gemini)",
     "option.timeout": "Espera máxima por respuesta del modelo (predeterminado: 10 minutos)",
     "option.noMinimize": "Mantener visible la ventana de Chrome",
     "option.debug": "Guardar diagnósticos del navegador",
@@ -467,7 +483,9 @@ const MESSAGES = Object.freeze({
     "login.savingProfile": "Cerrando Chrome y guardando el perfil dedicado...",
     "login.verified": "Inicio de sesión de {provider} verificado mediante una nueva conexión CDP.",
     "login.stillGuest": "{provider} sigue en modo invitado. Se volverá a abrir Chrome; completa el último paso de inicio de sesión/continuación.",
-    "model.chooseInBrowser": "{provider} ha iniciado sesión. Si quieres, selecciona manualmente en el navegador el modelo que deseas usar y después pulsa Enter aquí. Pulsa Enter directamente para conservar el modelo actual.",
+    "model.chooseInBrowser": "{provider} ha iniciado sesión.\nPuedes cambiar el modelo en el navegador.\nPulsa Enter para continuar o espera 5 segundos.",
+    "model.continueCountdown": "Continuando en {seconds}s…",
+    "model.continuing": "Continuando…",
     "model.setupCancelled": "Se canceló la confirmación de selección del modelo.",
     "task.youPrompt": "tú ›",
     "option.yes": "Omitir la confirmación",
@@ -536,6 +554,7 @@ const MESSAGES = Object.freeze({
     "option.project": "Répertoire du projet",
     "option.model": "Fournisseur d’IA web : {providers} (par défaut : {defaultProvider})",
     "option.once": "Exécuter une seule requête puis quitter",
+    "option.type": "Type de requête : agent (par défaut) ou music (Gemini uniquement)",
     "option.timeout": "Attente maximale d’une réponse du modèle (10 minutes par défaut)",
     "option.noMinimize": "Garder la fenêtre Chrome visible",
     "option.debug": "Écrire les diagnostics du navigateur",
@@ -553,7 +572,9 @@ const MESSAGES = Object.freeze({
     "login.savingProfile": "Fermeture de Chrome et enregistrement du profil dédié...",
     "login.verified": "Connexion à {provider} vérifiée via une nouvelle connexion CDP.",
     "login.stillGuest": "{provider} est encore en mode invité. Chrome va être rouvert ; terminez la dernière étape de connexion/continuation.",
-    "model.chooseInBrowser": "{provider} est connecté. Si nécessaire, choisissez manuellement dans le navigateur le modèle à utiliser, puis appuyez ici sur Entrée. Appuyez directement sur Entrée pour conserver le modèle actuel.",
+    "model.chooseInBrowser": "{provider} est connecté.\nVous pouvez changer de modèle dans le navigateur.\nAppuyez sur Entrée pour continuer ou attendez 5 secondes.",
+    "model.continueCountdown": "Reprise dans {seconds}s…",
+    "model.continuing": "Reprise…",
     "model.setupCancelled": "La confirmation du choix du modèle a été annulée.",
     "task.youPrompt": "vous ›",
     "option.yes": "Ignorer la demande de confirmation",
@@ -622,6 +643,7 @@ const MESSAGES = Object.freeze({
     "option.project": "Projektverzeichnis",
     "option.model": "Web-AI-Anbieter: {providers} (Standard: {defaultProvider})",
     "option.once": "Nur eine Anfrage ausführen und danach beenden",
+    "option.type": "Anfragetyp: agent (Standard) oder music (nur Gemini)",
     "option.timeout": "Maximale Wartezeit pro Modellantwort (Standard: 10 Minuten)",
     "option.noMinimize": "Chrome-Fenster sichtbar lassen",
     "option.debug": "Browserdiagnose schreiben",
@@ -639,7 +661,9 @@ const MESSAGES = Object.freeze({
     "login.savingProfile": "Chrome wird geschlossen und das dedizierte Profil gespeichert...",
     "login.verified": "{provider}-Anmeldung über eine neue CDP-Verbindung bestätigt.",
     "login.stillGuest": "{provider} befindet sich noch im Gastmodus. Chrome wird erneut geöffnet; schließe den letzten Anmelde-/Fortsetzen-Schritt ab.",
-    "model.chooseInBrowser": "{provider} ist angemeldet. Wähle bei Bedarf das gewünschte Modell manuell im Browser und drücke danach hier Enter. Drücke direkt Enter, um das aktuell ausgewählte Modell beizubehalten.",
+    "model.chooseInBrowser": "{provider} ist angemeldet.\nDu kannst das Modell im Browser ändern.\nDrücke Enter zum Fortfahren oder warte 5 Sekunden.",
+    "model.continueCountdown": "Weiter in {seconds}s…",
+    "model.continuing": "Weiter…",
     "model.setupCancelled": "Bestätigung der Modellauswahl abgebrochen.",
     "task.youPrompt": "du ›",
     "option.yes": "Bestätigungsabfrage überspringen",
@@ -708,6 +732,7 @@ const MESSAGES = Object.freeze({
     "option.project": "Diretório do projeto",
     "option.model": "Provedor de IA web: {providers} (padrão: {defaultProvider})",
     "option.once": "Executar apenas uma solicitação e sair",
+    "option.type": "Tipo de solicitação: agent (padrão) ou music (somente Gemini)",
     "option.timeout": "Espera máxima por uma resposta do modelo (padrão: 10 minutos)",
     "option.noMinimize": "Manter a janela do Chrome visível",
     "option.debug": "Gravar diagnósticos do navegador",
@@ -725,7 +750,9 @@ const MESSAGES = Object.freeze({
     "login.savingProfile": "Fechando o Chrome e salvando o perfil dedicado...",
     "login.verified": "Login de {provider} verificado por uma nova conexão CDP.",
     "login.stillGuest": "{provider} ainda está no modo visitante. O Chrome será reaberto; conclua a última etapa de login/continuação.",
-    "model.chooseInBrowser": "{provider} está conectado. Se quiser, escolha manualmente no navegador o modelo a usar e depois pressione Enter aqui. Pressione Enter imediatamente para manter o modelo atual.",
+    "model.chooseInBrowser": "{provider} está conectado.\nVocê pode alterar o modelo no navegador.\nPressione Enter para continuar ou aguarde 5 segundos.",
+    "model.continueCountdown": "Continuando em {seconds}s…",
+    "model.continuing": "Continuando…",
     "model.setupCancelled": "A confirmação da seleção do modelo foi cancelada.",
     "task.youPrompt": "você ›",
     "option.yes": "Ignorar a confirmação",
@@ -794,6 +821,7 @@ const MESSAGES = Object.freeze({
     "option.project": "Каталог проекта",
     "option.model": "Провайдер веб-ИИ: {providers} (по умолчанию: {defaultProvider})",
     "option.once": "Выполнить один запрос и завершить работу",
+    "option.type": "Тип запроса: agent (по умолчанию) или music (только Gemini)",
     "option.timeout": "Максимальное ожидание ответа модели (по умолчанию: 10 минут)",
     "option.noMinimize": "Не сворачивать окно Chrome",
     "option.debug": "Записывать диагностику браузера",
@@ -811,7 +839,9 @@ const MESSAGES = Object.freeze({
     "login.savingProfile": "Chrome закрывается, отдельный профиль сохраняется...",
     "login.verified": "Вход в {provider} подтверждён через новое соединение CDP.",
     "login.stillGuest": "{provider} всё ещё работает в гостевом режиме. Chrome будет открыт снова; завершите последний шаг входа/продолжения.",
-    "model.chooseInBrowser": "{provider}: вход выполнен. При необходимости вручную выберите нужную модель в браузере, затем нажмите Enter здесь. Чтобы оставить текущую модель, сразу нажмите Enter.",
+    "model.chooseInBrowser": "{provider}: вход выполнен.\nПри необходимости смените модель в браузере.\nНажмите Enter или подождите 5 секунд для продолжения.",
+    "model.continueCountdown": "Продолжение через {seconds} с…",
+    "model.continuing": "Продолжение…",
     "model.setupCancelled": "Подтверждение выбора модели отменено.",
     "task.youPrompt": "вы ›",
     "option.yes": "Пропустить запрос подтверждения",

package/src/cli/main.js

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import fs from "node:fs/promises";
 import path from "node:path";
-import { pathToFileURL } from "node:url";
+import { fileURLToPath } from "node:url";
 import { Command } from "commander";
 import { confirm, select } from "@inquirer/prompts";
 import { launchNativeLoginBrowser } from "../browser/native-login.js";
@@ -33,6 +33,8 @@ import { PolicyEngine } from "../policy/policy-engine.js";
 import { ApprovalStore } from "../policy/approval-store.js";
 import { createDefaultToolRegistry } from "../tools/default-tools.js";
 import { ProcessManager } from "../tools/process-manager.js";
+import { NativeImageReceiver } from "../image/native-image-receiver.js";
+import { NativeMusicReceiver, validateGenerationType } from "../audio/native-music-receiver.js";
 import { resolveLimits } from "../shared/limits.js";
 import { BrowserAdapterError } from "../shared/errors.js";
 import { EXPORTERS } from "../session/session-export.js";
@@ -41,6 +43,7 @@ import { extractAtMentions } from "./at-files.js";
 import {
   classifyChatInput,
   promptForConfirm,
+  promptForContinue,
   promptForText,
   promptForSelect,
   readChatMessage,
@@ -241,6 +244,9 @@ class ConversationRunner {
       // interactive TTY sessions where stdin is available to listen on.
       cancelOnEsc: interactive,
     });
+    const generationType = validateGenerationType(session.state.generationType, provider.id);
+    this.nativeMusicReceiver = generationType === "music" ? new NativeMusicReceiver({ adapter: this.adapter }) : null;
+    this.nativeImageReceiver = this.nativeMusicReceiver ? null : NativeImageReceiver.forAdapter(provider.id, this.adapter);
     this.interrupted = false;
     this.closed = false;
   }
@@ -248,6 +254,8 @@ class ConversationRunner {
   #buildRuntime() {
     return new AgentRuntime({
       adapter: this.adapter,
+      nativeImageReceiver: this.nativeImageReceiver,
+      nativeMusicReceiver: this.nativeMusicReceiver,
       registry: createDefaultToolRegistry({
         processManager: this.processManager,
         limits: this.limits,
@@ -261,10 +269,12 @@ class ConversationRunner {
             this.renderer.stopSpinner();
             await adapter.restoreWindow?.();
             try {
-              const answer = await promptForText({
+              const answer = await promptForContinue({
                 message: t("model.chooseInBrowser", {
                   provider: this.provider.label,
                 }),
+                countdown: (seconds) => t(seconds > 0
+                  ? "model.continueCountdown" : "model.continuing", { seconds }),
               });
               if (answer == null) {
                 throw new Error(t("model.setupCancelled"));
@@ -749,6 +759,8 @@ async function runAgent(taskParts, options) {
   // Fail fast on an unknown/unsupported --model before any startup work.
   const provider = resolveProvider(options.model ?? DEFAULT_PROVIDER);
 
+  const generationType = validateGenerationType(options.type, provider.id);
+
   const startup = await maybeRunStartupChecks(options);
   if (startup === "updated" || startup === "aborted") {
     return null;
@@ -789,6 +801,7 @@ async function runAgent(taskParts, options) {
     projectRoot,
     provider: provider.id,
     mode: null,
+    generationType,
   });
 
   // Resolve @file attachments in the opening task, if any. The task itself is
@@ -863,6 +876,9 @@ async function runResume(sessionId, instructionParts, options) {
   }
 
   const provider = resolveProvider(session.state.provider ?? DEFAULT_PROVIDER);
+  if (options.type != null && validateGenerationType(options.type, provider.id) !== (session.state.generationType ?? "agent")) {
+    throw new Error("--type cannot change the generation type of an existing session.");
+  }
   // A conversation belongs to one provider. --model on resume is only allowed
   // if it names the same provider; switching mid-conversation is rejected.
   if (options.model != null && getProvider(options.model).id !== provider.id) {
@@ -988,6 +1004,7 @@ const program = new Command()
     }),
   )
   .option("--once", t("option.once"), false)
+  .option("--type <type>", t("option.type"))
   .option("--model-turn-timeout-ms <milliseconds>", t("option.timeout"))
   .option("--no-minimize", t("option.noMinimize"))
   .option("--debug", t("option.debug"), false)
@@ -1055,7 +1072,13 @@ program
 const entryPath = process.argv[1]
   ? await fs.realpath(process.argv[1]).catch(() => null)
   : null;
-if (entryPath && import.meta.url === pathToFileURL(entryPath).href) {
+const modulePath = await fs.realpath(fileURLToPath(import.meta.url)).catch(() => null);
+// npm's Windows shim can use a short path or differently cased path. Compare
+// canonical filesystem paths rather than independently encoded file URLs.
+const isEntryPoint = entryPath && modulePath && (process.platform === "win32"
+  ? entryPath.toLowerCase() === modulePath.toLowerCase()
+  : entryPath === modulePath);
+if (isEntryPoint) {
   await program.parseAsync().catch((error) => {
     // Expected, actionable failures carry a plain message instead of a stack
     // trace (e.g. the provider's Chrome profile is locked by another session).
@@ -1065,6 +1088,12 @@ if (entryPath && import.meta.url === pathToFileURL(entryPath).href) {
       return;
     }
     console.error(error.stack ?? error.message);
+    if (error.details?.diagnosticFile) {
+      const dom = error.details.dom;
+      const explanation = dom?.issues?.length ? dom.issues.join(", ") : dom?.status ?? "unavailable";
+      console.error(`Page compatibility: ${explanation}`);
+      console.error(`Diagnostic report: ${error.details.diagnosticFile}`);
+    }
     process.exitCode = 1;
   }).finally(() => {
     // By the time every command finishes, all per-command cleanup (browser

package/src/cli/prompt-input.js

@@ -849,6 +849,74 @@ export async function promptForText(config, options = {}) {
   });
 }
 
+// A non-permission startup pause: Enter and expiry both continue. Keep this
+// separate from text/approval prompts, which must never accept on a timer.
+export async function promptForContinue({ message, countdown, timeoutMs = 5_000 }, {
+  inputStream = process.stdin,
+  outputStream = process.stdout,
+} = {}) {
+  if (inputStream.readableEnded || inputStream.destroyed) return null;
+  return await new Promise((resolve, reject) => {
+    const previousRaw = Boolean(inputStream.isRaw);
+    const deadline = Date.now() + timeoutMs;
+    let timer;
+    let settled = false;
+    const render = (seconds) => {
+      const text = countdown(seconds);
+      if (outputStream.isTTY) {
+        // Keep the changing status on one physical terminal row.
+        let visible = "";
+        for (const ch of text) {
+          if (displayWidth(visible + ch) >= (outputStream.columns || 80)) break;
+          visible += ch;
+        }
+        outputStream.write(`\r\u001b[2K${visible}`);
+      } else {
+        outputStream.write(`${text}\n`);
+      }
+    };
+    const finish = (value, error) => {
+      if (settled) return;
+      settled = true;
+      clearInterval(timer);
+      inputStream.removeListener("data", onData);
+      inputStream.removeListener("end", onEnd);
+      inputStream.removeListener("close", onEnd);
+      inputStream.removeListener("error", onError);
+      inputStream.setRawMode?.(previousRaw);
+      inputStream.pause?.();
+      if (value) render(0);
+      if (outputStream.isTTY) outputStream.write("\n");
+      if (error) reject(error);
+      else resolve(value);
+    };
+    const onData = (chunk) => {
+      const text = String(chunk);
+      if (/[\u0003\u0004]/.test(text)) finish(null);
+      else if (/[\r\n]/.test(text)) finish(true);
+    };
+    const onEnd = () => finish(null);
+    const onError = (error) => finish(null, error);
+    inputStream.on("data", onData);
+    inputStream.once("end", onEnd);
+    inputStream.once("close", onEnd);
+    inputStream.once("error", onError);
+    try {
+      inputStream.setRawMode?.(true);
+      outputStream.write(`${message}\n`);
+      render(Math.ceil(timeoutMs / 1_000));
+      timer = setInterval(() => {
+        const seconds = Math.max(0, Math.ceil((deadline - Date.now()) / 1_000));
+        if (seconds === 0) finish(true);
+        else render(seconds);
+      }, 1_000);
+      inputStream.resume?.();
+    } catch (error) {
+      finish(null, error);
+    }
+  });
+}
+
 export async function promptForSelect(config, options = {}) {
   return await promptWithCleanExit(config, {
     ...options,