feat: login recipe auto-discovery (discover command)

- discover <source>: open homepage → find login entry by visible text
  (fallback to common login paths) → click/open → detect main-frame vs
  iframe form (captcha/challenge iframes excluded) → enumerate fields →
  generate recipe (single-page / two-step / password-only) with
  remember-me checkbox ticking → write logins/<source>.json → optional
  live login verification (--no-login to skip)
- loggedOutText: login-state check by visible button/link text
  (regex), for sites whose login control has no stable class
- honest boundaries: CAPTCHA/OTP/perimeter protection exits with a
  clear human-required message and writes nothing (validated against
  wsj's DataDome layer); validated green path against nikkei Piano ID

skills/news-harvester/scripts/login-tools.js

@@ -22,7 +22,14 @@
  *
  * 浏览器实例:尊重 CDP_PORT_OVERRIDE / CDP_PROFILE_OVERRIDE(与 ensure-chromium 一致),
  * 未设置时用 config.chromium.cdpPort(默认 9222)。
- * 退出码:0 成功;1 环境/配方错误;2 浏览器不可用;3 未登录;4 无配方;5 登录失败。
+ * discover 子命令(自动配置配方):
+ *   node scripts/login-tools.js discover <source> [--no-login]
+ *   打开站点主页 → 按文本找登录入口 → 识别弹层/跳转/iframe → 枚举表单生成配方
+ *   → 写入 logins/<source>.json → (默认)立即用凭据实登验证。
+ *   站点主页取 NH_LOGIN_URL 环境变量,缺省读技能 config 的 sources[].homepageUrl。
+ *   识别不了的形态(CAPTCHA/OTP/OAuth)会如实报告,不写配方。
+ *
+ * 退出码:0 成功;1 环境/配方错误;2 浏览器不可用;3 未登录;4 无配方;5 登录失败;6 自动发现失败。
  */
 
 const fs = require('fs');
@@ -143,18 +150,143 @@ async function closeTab(id) {
   await fetch(`http://127.0.0.1:${cdpPort()}/json/close/${id}`, { method: 'PUT' }).catch(() => {});
 }
 
+async function homepageOf(source) {
+  if (process.env.NH_LOGIN_URL) return process.env.NH_LOGIN_URL;
+  try {
+    const cfg = JSON.parse(fs.readFileSync(path.join(ROOT, 'config.json'), 'utf8'));
+    const s = (cfg.sources || []).find((x) => x.id === source);
+    if (s && s.homepageUrl) return s.homepageUrl;
+  } catch (e) { /* fallthrough */ }
+  return null;
+}
+
+const LOGIN_ENTRY_RE = 'log\\s*in|sign\\s*in|登\\s*入|登录|登陸';
+
+// ── discover:自动发现登录流程并生成配方 ──────────────────────
+async function discover(source, doLogin) {
+  const home = await homepageOf(source);
+  if (!home) { console.error('找不到站点主页:设 NH_LOGIN_URL 或在技能 config 配置 homepageUrl'); process.exit(6); }
+  console.log(`① 打开主页: ${home}`);
+  const opened = await openTab(home);
+  if (!opened) { console.error('无法打开主页'); process.exit(6); }
+  await sleep(6000);
+  let tab = await tabById(opened.id);
+  if (!tab) { console.error('主页标签丢失'); process.exit(6); }
+
+  // ② 找登录入口并点击(按可见文本)
+  console.log('② 查找登录入口…');
+  let entry = await evalIn(tab, `(() => {
+    const els = [...document.querySelectorAll('button,[role=button],a')];
+    const el = els.find(e => new RegExp(${JSON.stringify(LOGIN_ENTRY_RE)}, 'i').test((e.innerText || '').trim()) && e.offsetParent);
+    if (!el) return null;
+    const text = (el.innerText || '').trim();
+    el.click();
+    return { text, tag: el.tagName };
+  })()`);
+  if (!entry) {
+    // 回退:常见登录路径探测
+    for (const p of ['/login', '/signin', '/register/signin.php', '/user/login']) {
+      const probe = await openTab(new URL(p, new URL(home).origin).href);
+      if (!probe) continue;
+      await sleep(4000);
+      const hasForm = await evalIn(await tabById(probe.id) || probe,
+        `!!document.querySelector('input[type=password], input[type=email], input[name*=mail i]')`);
+      if (hasForm) { tab = await tabById(probe.id); entry = { text: '', tag: 'A', via: p }; break; }
+      await closeTab(probe.id).catch(() => {});
+    }
+  }
+  if (!entry) { console.error('未找到登录入口(可能已登录/需人工识别)'); process.exit(6); }
+  console.log(`   入口: ${entry.text || entry.via}(${entry.tag})`);
+  await sleep(5000);
+  tab = (await tabById(opened.id)) || tab;
+
+  // ③ 判断表单上下文:主框架 or iframe(排除反爬/验证码 iframe)
+  let frameSub = null;
+  const iframeGuess = await evalIn(tab, `(() => {
+    const bad = /captcha|challenge|delivery|turnstile|recaptcha|hcaptcha|perimeter|px-|datadome/i;
+    const good = /auth|login|signin|sign-in|account|sso/i;
+    const frames = [...document.querySelectorAll('iframe')].map(f => f.src || '').filter(s => s);
+    const hit = frames.find(s => good.test(s) && !bad.test(s)) || null;
+    return hit ? { src: hit.slice(0, 110), blocked: frames.some(s => bad.test(s)) && !hit } : { src: null, blocked: frames.some(s => bad.test(s)) };
+  })()`);
+  if (iframeGuess && iframeGuess.src) {
+    frameSub = new URL(iframeGuess.src).host;
+    console.log(`   登录表单在 iframe: ${frameSub}`);
+  } else if (iframeGuess && iframeGuess.blocked) {
+    console.error('检测到反爬/验证码层(captcha/challenge iframe),无法自动配置,需人工登录一次');
+    process.exit(6);
+  } else {
+    console.log('   登录表单在主框架');
+  }
+
+  // ④ 枚举表单(在对应上下文里执行)
+  const dumpExpr = `JSON.stringify({
+    email: (() => { const e = document.querySelector('input[type=email]') || document.querySelector('input[name*=mail i]') || document.querySelector('input[name=email i]'); return e ? (e.id ? '#' + e.id : (e.name ? '[name=' + JSON.stringify(e.name) + ']' : 'input[type=email]')) : null; })(),
+    emailVisible: (() => { const e = document.querySelector('input[type=email],input[name*=mail i],input[name=email i]'); return e ? e.offsetParent !== null : false; })(),
+    pw: !!document.querySelector('input[type=password]'),
+    pwVisible: (() => { const e = document.querySelector('input[type=password]'); return e ? e.offsetParent !== null : false; })(),
+    pwSel: (() => { const e = document.querySelector('input[type=password]'); return e ? 'input[type=password]' : null; })(),
+    submit: (() => { const b = document.querySelector('form button[type=submit], form input[type=submit], button[type=submit]'); return b ? 'button[type=submit], form input[type=submit]' : null; })(),
+    boxes: [...document.querySelectorAll('input[type=checkbox]')].map(c => ((c.closest('label') || {}).innerText || '').replace(/\\s+/g, ' ').slice(0, 30)).filter(Boolean),
+    captcha: !!document.querySelector('.g-recaptcha, iframe[src*=recaptcha], iframe[src*=hcaptcha], iframe[src*=turnstile]')
+  })`;
+  let info = {};
+  try {
+    info = JSON.parse((await evalIn(tab, dumpExpr, frameSub)) || '{}');
+  } catch (e) {
+    if (frameSub) {
+      console.log(`   iframe(${frameSub})无执行上下文,回退主框架…`);
+      info = JSON.parse((await evalIn(tab, dumpExpr)) || '{}');
+    } else { throw e; }
+  }
+  if (info.captcha) { console.error('检测到 CAPTCHA,无法自动配置(需人工)'); process.exit(6); }
+  if (!info.email && !info.pw) { console.error('未发现登录表单字段(可能是 OAuth/二维码登录)'); process.exit(6); }
+
+  // ⑤ 生成配方
+  const steps = [];
+  if (entry.text && !entry.via) steps.push({ click: { text: entry.text } });
+  if (info.email && (!info.pwVisible)) {
+    steps.push({ ...(frameSub ? { frame: frameSub } : {}), fill: { [info.email]: '$USER' }, submit: info.submit });
+    steps.push({ ...(frameSub ? { frame: frameSub } : {}), fill: { 'input[type=password]': '$PASS' }, submit: info.submit, waitMs: 8000 });
+    console.log('⑤ 识别为两步式(邮箱 → 密码)');
+  } else if (info.email && info.pw) {
+    const fill = { [info.email]: '$USER', 'input[type=password]': '$PASS' };
+    steps.push({ ...(frameSub ? { frame: frameSub } : {}), fill, submit: info.submit, waitMs: 8000 });
+    console.log('⑤ 识别为单页式(邮箱+密码)');
+  } else if (info.pw) {
+    steps.push({ ...(frameSub ? { frame: frameSub } : {}), fill: { 'input[type=password]': '$PASS' }, submit: info.submit, waitMs: 8000 });
+    console.log('⑤ 仅见密码框(账号可能已被站点记住)');
+  }
+  if (info.boxes.length) steps[steps.length - 1].checks = [{ selector: 'input[type=checkbox]', where: 'save|keep|remember|signed|记住|保持' }];
+  const recipe = {
+    checkUrl: home,
+    ...(entry.text ? { loggedOutText: entry.text.replace(/[.*+?^${}()|[\]\\]/g, '\\$&') } : {}),
+    steps,
+  };
+  fs.mkdirSync(path.join(ROOT, 'logins'), { recursive: true });
+  fs.writeFileSync(path.join(ROOT, 'logins', `${source}.json`), JSON.stringify(recipe, null, 2) + '\n');
+  console.log(`⑥ 配方已写入 logins/${source}.json`);
+  if (!doLogin) { console.log('⑦ --no-login:跳过实登验证'); process.exit(0); }
+}
+
 async function main() {
-  const [command, source] = process.argv.slice(2);
+  const [command, source, ...rest] = process.argv.slice(2);
   if (!command || !source) {
-    console.error('用法: node scripts/login-tools.js check|login <source>');
+    console.error('用法: node scripts/login-tools.js check|login|discover <source>');
     process.exit(1);
   }
+  if (command === 'discover') {
+    if (!(await browserAlive())) { console.error(`浏览器实例不可用(端口 ${cdpPort()})`); process.exit(2); }
+    if (!LOGIN_USER || !LOGIN_PASS) { console.error('缺少凭据: 需要环境变量 NH_LOGIN_USER / NH_LOGIN_PASS'); process.exit(1); }
+    await discover(source, !rest.includes('--no-login'));
+    // discover 写好配方后,直接走 login 分支完成实登验证
+  }
   const recipe = loadRecipe(source);
   if (!recipe) {
     console.error(`无登录配方: logins/${source}.json(该源无需登录或未配置)`);
     process.exit(4);
   }
-  if (command === 'login' && (!LOGIN_USER || !LOGIN_PASS)) {
+  if ((command === 'login' || command === 'discover') && (!LOGIN_USER || !LOGIN_PASS)) {
     console.error('缺少凭据: 需要环境变量 NH_LOGIN_USER / NH_LOGIN_PASS');
     process.exit(1);
   }
@@ -171,6 +303,10 @@ async function main() {
     if (recipe.loggedOutSelector) {
       return !(await evalIn(tab, `!!document.querySelector(${JSON.stringify(recipe.loggedOutSelector)})`));
     }
+    if (recipe.loggedOutText) {
+      const re = recipe.loggedOutText;
+      return !(await evalIn(tab, `!![...document.querySelectorAll('button,[role=button],a')].find(e => new RegExp(${JSON.stringify(re)}, 'i').test((e.innerText || '').trim()) && e.offsetParent)`));
+    }
     return false;
   };
 
@@ -186,8 +322,8 @@ async function main() {
     process.exit(logged ? 0 : 3);
   }
 
-  // ── login:按配方多步执行 ──
-  if (command === 'login') {
+  // ── login:按配方多步执行(discover 成功后落入此处实登验证) ──
+  if (command === 'login' || command === 'discover') {
     let ownTabId = null;  // 全程复用同一个标签(提交导航不换 target id)
     for (const step of recipe.steps) {
       if (step.url) {