@xbrowser/cli 1.11.0 → 1.13.0

package/dist/{cdp-driver-GD6YBBGE.js → cdp-driver-HYCLE4YP.js}

@@ -14,7 +14,7 @@ import {
   scrollIntoView,
   waitForActionable,
   waitForNetworkIdle
-} from "./chunk-7OFM755Z.js";
+} from "./chunk-R2POF2GP.js";
 import {
   connectToCDP,
   findChrome,
@@ -23,6 +23,7 @@ import {
   launchChrome
 } from "./chunk-TNEN6VQ2.js";
 import "./chunk-GDKLH7ZY.js";
+import "./chunk-H2A5JUK5.js";
 import "./chunk-KFQGP6VL.js";
 export {
   CDPConnection,

package/dist/{cdp-driver-VEK6BNN6.js → cdp-driver-VS7Z7W7Y.js}

@@ -14,7 +14,7 @@ import {
   scrollIntoView,
   waitForActionable,
   waitForNetworkIdle
-} from "./chunk-XQ4HRPDJ.js";
+} from "./chunk-T2CNKWPV.js";
 import "./chunk-3FWLW7FS.js";
 import {
   connectToCDP,

package/dist/chunk-A6LPGFAL.js

@@ -0,0 +1,437 @@
+import {
+  createRuleEngine
+} from "./chunk-H2A5JUK5.js";
+
+// src/cdp-interceptor/proxy.ts
+import { WebSocketServer, WebSocket } from "ws";
+
+// src/cdp-interceptor/logger.ts
+function createLogger(config) {
+  const buffer = [];
+  const MAX_BUFFER = 2e3;
+  return {
+    info(message, meta) {
+      if (!config.enableLogging) return;
+      const ts = (/* @__PURE__ */ new Date()).toISOString();
+      if (meta) {
+        console.log(`[CDPInterceptor ${ts}] ${message}`, JSON.stringify(meta));
+      } else {
+        console.log(`[CDPInterceptor ${ts}] ${message}`);
+      }
+    },
+    log(method, direction, sessionId, payload, decision) {
+      const entry = {
+        timestamp: Date.now(),
+        direction,
+        sessionId,
+        method,
+        payload: sanitizePayload(payload),
+        decision: decision ?? void 0
+      };
+      if (config.enableLogging) {
+        buffer.push(entry);
+        if (buffer.length > MAX_BUFFER) buffer.shift();
+        const tag = decision ? decision.action === "block" ? "\u{1F6AB}BLOCK" : decision.action === "transform" ? "\u{1F504}XFMR" : "\u2705" : "  ";
+        const reason = decision ? ` [${decision.severity}] ${decision.reason}` : "";
+        console.log(`[CDP] ${tag} ${direction} ${method}${reason}`);
+      }
+      return entry;
+    },
+    getRecent(count) {
+      return buffer.slice(-count);
+    },
+    flush() {
+      buffer.length = 0;
+    }
+  };
+}
+function sanitizePayload(payload) {
+  if (typeof payload !== "object" || payload === null) return { raw: String(payload) };
+  const obj = payload;
+  const cleaned = {};
+  for (const [key, value] of Object.entries(obj)) {
+    if (key === "data" && typeof value === "string" && value.length > 200) {
+      cleaned[key] = `<binary: ${value.length} chars>`;
+    } else if (key === "expression" && typeof value === "string" && value.length > 500) {
+      cleaned[key] = value.substring(0, 500) + "...";
+    } else {
+      cleaned[key] = value;
+    }
+  }
+  return cleaned;
+}
+
+// src/cdp-interceptor/advisor.ts
+function formatBlockMessage(decision, method) {
+  const adv = advise(decision, method);
+  const lines = [
+    `[CDP-FIREWALL-BLOCK] rule=${decision.ruleId}`,
+    `method=${method}`,
+    `reason=${decision.reason}`,
+    `suggestion=${decision.suggestion ?? adv.detail}`
+  ];
+  if (adv.codeExample) {
+    lines.push(`code-example=`);
+    lines.push(adv.codeExample);
+  }
+  return lines.join("\n");
+}
+function advise(decision, originalMethod) {
+  const baseAdvice = getBaseAdvice(decision, originalMethod);
+  return {
+    ruleId: decision.ruleId,
+    title: baseAdvice.title,
+    detail: decision.suggestion ?? baseAdvice.detail,
+    codeExample: baseAdvice.codeExample
+  };
+}
+function getBaseAdvice(decision, method) {
+  switch (decision.ruleId) {
+    case "dom-mutation":
+      return {
+        title: "Direct DOM property mutation blocked",
+        detail: `Your ${method} call tried to set a DOM property directly. In React/Vue/Angular, this bypasses the framework's virtual DOM completely meaning onChange/onInput never fires. The website CAN detect this mismatch as automation.`,
+        codeExample: [
+          "# \u274C BLOCKED \u2014 what you tried to do:",
+          `page.evaluate(\`el.value = 'hello'\`)  # triggers isTrusted=false`,
+          "",
+          "# \u2705 USE INSTEAD \u2014 proper CDP input dispatch:",
+          "page.fill('#selector', 'hello')           # Playwright: dispatches input+change events",
+          "page.type('#selector', 'hello', {delay})  # Playwright: real keystrokes",
+          "page.locator('#selector').fill('hello')   # Playwright: recommended API"
+        ].join("\n")
+      };
+    case "mouse-trajectory":
+      return {
+        title: "Unnatural mouse trajectory blocked",
+        detail: `Your ${method} sequence formed a perfectly linear path. Human hands have micro-tremors (1-3px variation at any point along the arc), acceleration curves, and never draw straight lines between distant points.`,
+        codeExample: [
+          "# \u274C BLOCKED \u2014 linear interpolation:",
+          "for i in range(20):",
+          "    page.mouse.move(x0 + (x1-x0)*(i/20), y0 + (y1-y0)*(i/20))",
+          "",
+          "# \u2705 USE INSTEAD \u2014 Bezier curves with overshoot:",
+          '# Use "ghost-cursor" or similar library',
+          "from ghost_cursor import path_to",
+          "path_to(page, (x1, y1))"
+        ].join("\n")
+      };
+    case "input-keystroke":
+      if (method === "Input.insertText") {
+        return {
+          title: "Input.insertText detected (logged, not blocked)",
+          detail: "Input.insertText bypasses native keyDown\u2192keyPress\u2192input\u2192keyUp events. Playwright uses this internally for page.fill(). Logged for observability."
+        };
+      }
+      return {
+        title: "Unnatural keystroke timing blocked",
+        detail: `Your ${method} calls have unnaturally constant timing (e.g., exact 50ms intervals). Human typing always has variation (CV > 0.2).`,
+        codeExample: [
+          "# \u274C BLOCKED \u2014 exact constant delay:",
+          "page.type('#input', 'hello', {delay: 50})  # every keystroke exactly 50ms apart",
+          "",
+          "# \u2705 USE INSTEAD \u2014 variable delay (human-like):",
+          "page.fill('#input', 'hello')               # recommended, dispatches events properly",
+          "# OR: type with randomized delay",
+          "page.type('#input', 'hello', {delay: 50 + Math.floor(Math.random() * 80)})"
+        ].join("\n")
+      };
+    case "automation-signals":
+      return {
+        title: "Browser automation marker detected",
+        detail: `Your ${method} call accessed a property/marker that anti-crawler systems check to detect automation. These markers (navigator.webdriver, window.__playwright, etc.) are the #1 detection vector.`,
+        codeExample: [
+          "# \u274C BLOCKED \u2014 don't check for automation markers:",
+          "navigator.webdriver      # NEVER check this",
+          "window.__playwright      # NEVER check this",
+          "chrome.runtime           # NEVER check this",
+          "",
+          "# \u2705 Just go about your business normally.",
+          "# Anti-detection is handled by the CDP firewall automatically."
+        ].join("\n")
+      };
+    case "fingerprinting":
+      return {
+        title: "Browser fingerprinting API access blocked",
+        detail: `Your ${method} call accessed a known fingerprinting API. These APIs (canvas.toDataURL, WebGL getParameter, AudioContext, etc.) are used by anti-crawler systems to build a unique device fingerprint.`,
+        codeExample: [
+          "# \u274C BLOCKED \u2014 fingerprinting vector:",
+          "canvas.toDataURL()                        # returns unique browser hash",
+          'gl.getParameter(gl.VENDOR)                # returns "SwiftShader" in headless',
+          "screen.availWidth - screen.availHeight    # no OS chrome in headless",
+          "",
+          "# \u2705 Avoid accessing these APIs. They are only used for fingerprinting."
+        ].join("\n")
+      };
+    case "event-simulation":
+      return {
+        title: "Synthetic event simulation blocked",
+        detail: `Your ${method} call simulated user interaction via el.click() or dispatchEvent(new Event(...)). These produce isTrusted=false events, which are 100% detectable by any anti-crawler that checks isTrusted on critical events.`,
+        codeExample: [
+          "# \u274C BLOCKED \u2014 synthetic events (isTrusted=false):",
+          "el.click()                                 # isTrusted=false",
+          'el.dispatchEvent(new Event("click"))       # isTrusted=false',
+          "el.focus()                                 # isTrusted=false",
+          "",
+          "# \u2705 USE INSTEAD \u2014 CDP-level input dispatch (isTrusted=true):",
+          "page.click(selector)                       # uses Input.dispatchMouseEvent",
+          "page.fill(selector, value)                 # dispatches real input events"
+        ].join("\n")
+      };
+    case "emulation-override":
+      return {
+        title: "CDP emulation override blocked",
+        detail: `Your ${method} call overrides browser behavior in a way that creates detectable inconsistencies. Anti-crawler systems cross-check multiple sources (e.g., navigator.userAgent vs HTTP User-Agent header) to catch these mismatches.`,
+        codeExample: [
+          "# \u274C BLOCKED \u2014 detectable emulation override:",
+          "Emulation.setUserAgentOverride(...)        # JS vs HTTP header mismatch",
+          "Emulation.setGeolocationOverride(...)       # IP geo vs overridden geo mismatch",
+          "Emulation.setDeviceMetricsOverride(...)     # matchMedia vs actual viewport",
+          "",
+          "# \u2705 These are handled automatically by the CDP firewall.",
+          "# Do NOT call them manually."
+        ].join("\n")
+      };
+    case "network-anomaly":
+      return {
+        title: "Network anomaly detected",
+        detail: `Your ${method} call triggered a network pattern that is characteristic of scrapers: blocking URLs, clearing caches, or intercepting requests.`,
+        codeExample: [
+          "# \u274C BLOCKED \u2014 scraper optimization:",
+          "Network.clearBrowserCache()                # natural users never do this",
+          'Network.setBlockedURLs(["*fonts*"])        # blocking resources is detectable',
+          "Fetch.enable()                             # MITM-style interception",
+          "",
+          "# \u2705 Let the browser manage its own cache and network normally."
+        ].join("\n")
+      };
+    case "page-lifecycle":
+      return {
+        title: "Suspicious page lifecycle pattern blocked",
+        detail: `Your ${method} call reveals an unnatural page interaction sequence: navigating too fast, taking screenshots before the page renders, or generating PDFs (a telltale scraper giveaway).`,
+        codeExample: [
+          "# \u274C BLOCKED \u2014 unnatural lifecycle:",
+          "page.goto(url); page.pdf()                 # PDF = scraper giveaway",
+          "page.goto(url); page.screenshot() <500ms   # screenshot before render",
+          "page.goto(url) 3x in <100ms               # rapid navigation barrage",
+          "",
+          "# \u2705 Add proper waits between actions:",
+          'page.goto(url, {waitUntil: "networkidle"})',
+          'page.waitForSelector("body")',
+          "page.screenshot()                          # after rendering"
+        ].join("\n")
+      };
+    default:
+      return {
+        title: decision.reason,
+        detail: decision.suggestion ?? `The CDP call "${method}" was blocked by rule "${decision.ruleId}".`
+      };
+  }
+}
+
+// src/cdp-interceptor/proxy.ts
+function makeCompoundId(cdpSessionId, rawSessionId) {
+  return `${cdpSessionId ?? "nil"}::${rawSessionId ?? "nil"}`;
+}
+var CDPInterceptorProxy = class {
+  wss = null;
+  engine;
+  config;
+  logger;
+  started = false;
+  stats = {
+    totalMessages: 0,
+    blockedMessages: 0,
+    transformedMessages: 0,
+    passedMessages: 0,
+    byRule: {}
+  };
+  constructor(config) {
+    this.config = config;
+    this.engine = createRuleEngine(config.rules);
+    this.logger = createLogger({
+      enableLogging: config.enableLogging ?? true,
+      logDir: config.logDir
+    });
+  }
+  /** The port the proxy is listening on (only valid after start()) */
+  get port() {
+    const addr = this.wss?.address();
+    if (addr && typeof addr === "object") return addr.port;
+    return 0;
+  }
+  /** Start the proxy server */
+  async start() {
+    if (this.started) return this.port;
+    return new Promise((resolve, reject) => {
+      this.wss = new WebSocketServer({ port: this.config.listenPort ?? 0 }, () => {
+        const port = this.port;
+        this.engine.start();
+        this.started = true;
+        this.logger.info("CDP interceptor proxy started", { port, endpoint: this.config.cdpEndpoint });
+        resolve(port);
+      });
+      this.wss.on("error", reject);
+      this.wss.on("connection", (clientWs, _req) => {
+        this.handleConnection(clientWs);
+      });
+    });
+  }
+  /** Stop the proxy server */
+  async stop() {
+    this.engine.stop();
+    this.logger.flush();
+    this.started = false;
+    return new Promise((resolve) => {
+      if (!this.wss) return resolve();
+      this.wss.close(() => resolve());
+      this.wss = null;
+    });
+  }
+  /** Get accumulated statistics */
+  getStats() {
+    return { ...this.stats };
+  }
+  /** Get recent log entries (for inspection) */
+  getRecentLogs(count = 50) {
+    return this.logger.getRecent(count);
+  }
+  // ── Connection handling ──────────────────────────────────────
+  handleConnection(clientWs) {
+    let browserWs = null;
+    let isAlive = true;
+    const pendingMessages = [];
+    browserWs = new WebSocket(this.config.cdpEndpoint);
+    clientWs.on("message", (raw) => {
+      if (browserWs && browserWs.readyState === WebSocket.OPEN) {
+        this.handleClientMessage(clientWs, browserWs, raw);
+      } else {
+        pendingMessages.push(raw);
+      }
+    });
+    browserWs.on("open", () => {
+      for (const buf of pendingMessages) {
+        this.handleClientMessage(clientWs, browserWs, buf);
+      }
+      pendingMessages.length = 0;
+    });
+    browserWs.on("error", (err) => {
+      this.logger.info("Browser WebSocket error", { error: String(err) });
+    });
+    browserWs.on("close", (code, reason) => {
+      if (isAlive && clientWs.readyState === WebSocket.OPEN) {
+        this.logger.info("Browser WS closed, closing client", { code, reason: String(reason) });
+        clientWs.close();
+      }
+    });
+    browserWs.on("message", (raw) => {
+      this.handleBrowserMessage(clientWs, browserWs, raw);
+    });
+    const cleanup = () => {
+      isAlive = false;
+      if (browserWs && browserWs.readyState === WebSocket.OPEN) {
+        browserWs.close();
+      }
+    };
+    clientWs.on("close", cleanup);
+    clientWs.on("error", cleanup);
+    browserWs.on("close", () => {
+      if (isAlive && clientWs.readyState === WebSocket.OPEN) {
+        clientWs.close();
+      }
+    });
+    browserWs.on("error", cleanup);
+  }
+  // ── Message processing ───────────────────────────────────────
+  handleClientMessage(clientWs, browserWs, raw) {
+    const msg = this.parseMessage(raw);
+    if (!msg) return;
+    this.stats.totalMessages++;
+    if (!("method" in msg)) {
+      browserWs.send(raw.toString());
+      return;
+    }
+    const request = msg;
+    const ctx = {
+      method: request.method,
+      params: request.params ?? {},
+      sessionId: makeCompoundId("_cdpSession" in browserWs ? browserWs._cdpSession : void 0, request.sessionId),
+      direction: "client\u2192browser"
+    };
+    const decision = this.engine.evaluate(ctx);
+    this.logger.log(ctx.method, "client\u2192browser", ctx.sessionId, { method: ctx.method, params: ctx.params }, decision);
+    if (decision) {
+      this.recordDecision(decision);
+    }
+    if (decision?.action === "block") {
+      const blockMsg = formatBlockMessage(decision, ctx.method);
+      const errorResponse = {
+        id: request.id,
+        error: {
+          code: decision.errorCode ?? -32e3,
+          message: blockMsg
+        },
+        sessionId: request.sessionId
+      };
+      this.stats.blockedMessages++;
+      console.error(`
+${blockMsg}
+`);
+      clientWs.send(JSON.stringify(errorResponse));
+      return;
+    }
+    if (decision?.action === "transform" && decision.transformedParams) {
+      const transformed = { ...request, params: decision.transformedParams };
+      this.stats.transformedMessages++;
+      browserWs.send(JSON.stringify(transformed));
+      return;
+    }
+    this.stats.passedMessages++;
+    browserWs.send(raw.toString());
+  }
+  handleBrowserMessage(clientWs, _browserWs, raw) {
+    const msg = this.parseMessage(raw);
+    if (!msg) {
+      clientWs.send(raw.toString());
+      return;
+    }
+    if ("method" in msg) {
+      const event = msg;
+      const ctx = {
+        method: event.method,
+        params: event.params ?? {},
+        sessionId: event.sessionId ?? "browser",
+        direction: "browser\u2192client"
+      };
+      const decision = this.engine.evaluate(ctx);
+      if (decision?.action === "block") {
+        return;
+      }
+    }
+    clientWs.send(raw.toString());
+  }
+  // ── Utilities ────────────────────────────────────────────────
+  parseMessage(raw) {
+    try {
+      return JSON.parse(raw.toString());
+    } catch {
+      return null;
+    }
+  }
+  recordDecision(decision) {
+    if (!this.stats.byRule[decision.ruleId]) {
+      this.stats.byRule[decision.ruleId] = { matched: 0, blocked: 0, transformed: 0 };
+    }
+    this.stats.byRule[decision.ruleId].matched++;
+    if (decision.action === "block") {
+      this.stats.byRule[decision.ruleId].blocked++;
+    } else if (decision.action === "transform") {
+      this.stats.byRule[decision.ruleId].transformed++;
+    }
+  }
+};
+
+export {
+  advise,
+  CDPInterceptorProxy
+};