Merge pull request #2721 from alectimison-maker/fix/captcha-additive-matcher

fix(captcha): make English matcher additive

Author: esokullu

src/chrome/src/agent/agent.js

@@ -52,7 +52,7 @@ import { normalizeRuntimeTraceConfig } from '../trace/runtime-config.js';
 import { tracesToMarkdown } from './trace-export.js';
 import { solveCaptcha, detectCaptcha, injectToken, captchaParamError, captchaTypesMatch, captchaWebsiteUrl } from './captcha-solver.js';
 import { isCapsolverEnabled, normalizeCapsolverApiKey } from './capsolver-config.js';
-import { captchaChallengeKey, detectChallengeDialog, detectChallengeDialogInPage } from './captcha-gate.js';
+import { captchaChallengeKey, captchaChallengeMatcherOptions, detectChallengeDialog, detectChallengeDialogInPage } from './captcha-gate.js';
 import { applyCaptchaFrameVisibility } from './captcha-frame-runtime.js';
 import { getRecordingStateFresh as recorderStateFresh } from '../recorder/host.js';
 import { Capability, CAPABILITY_LABEL, capabilitiesFor, requiredHosts, frameHostMatches, isNetworkMutation, normalizeHost, PermissionManager, UNTRUSTED_CONTENT_TOOLS } from './permission-gate.js';
@@ -4205,7 +4205,7 @@ Rules: no prose intro, no conclusion, no "this screenshot shows...", no layout d
       func: detectChallengeDialogInPage,
       args: [{
         includeFrameContext: true,
-        allowGenericFailure: options?.allowGenericFailure === true,
+        ...captchaChallengeMatcherOptions(),
       }],
     });
     const inspected = (results) => {
@@ -4218,7 +4218,6 @@ Rules: no prose intro, no conclusion, no "this screenshot shows...", no layout d
         if (!label) return [];
         const normalized = detectChallengeDialog(
           `dialog ${JSON.stringify(label.slice(0, 200))}`,
-          { allowGenericFailure: options?.allowGenericFailure === true },
         );
         if (!normalized?.normalizedLabel) return [];
         return [{
@@ -4323,13 +4322,10 @@ Rules: no prose intro, no conclusion, no "this screenshot shows...", no layout d
       ? toolResult.captchaChallengeFrameId
       : null;
     let observedChallengeFrameUrl = String(toolResult.captchaChallengeFrameUrl || '');
-    let challenge = detectChallengeDialog(toolResult.pageContent, {
-      allowGenericFailure: !!activeGate,
-    });
+    let challenge = detectChallengeDialog(toolResult.pageContent);
     if (!challenge && toolResult.pageGate?.surface === 'dialog' && toolResult.pageGate?.label) {
       challenge = detectChallengeDialog(
         `dialog ${JSON.stringify(String(toolResult.pageGate.label).slice(0, 200))}`,
-        { allowGenericFailure: !!activeGate },
       );
     }
     if (
@@ -4340,12 +4336,10 @@ Rules: no prose intro, no conclusion, no "this screenshot shows...", no layout d
       const recheck = await this._detectChallengeDialogBeforeMutation(tabId, {
         includeStatus: true,
         expectedFrameId: observedChallengeFrameId,
-        allowGenericFailure: !!activeGate,
       });
       if (recheck.challenge?.label) {
         challenge = detectChallengeDialog(
           `dialog ${JSON.stringify(String(recheck.challenge.label).slice(0, 200))}`,
-          { allowGenericFailure: !!activeGate },
         );
         observedChallengeFrameId = Number.isInteger(recheck.challenge.frameId)
           ? recheck.challenge.frameId
@@ -4412,20 +4406,67 @@ Rules: no prose intro, no conclusion, no "this screenshot shows...", no layout d
         }
       }
     }
+    const requestedPage = toolArgs?.page;
+    const requestedMaxDepth = toolArgs?.maxDepth;
+    const parsedMaxDepth = Number(requestedMaxDepth);
+    const authoritativeRootRead = !toolArgs?.ref_id
+      && (
+        requestedPage === undefined
+        || requestedPage === null
+        || requestedPage === ''
+        || Number(requestedPage) === 1
+      )
+      && treeFilter !== 'interactive'
+      && (
+        requestedMaxDepth === undefined
+        || requestedMaxDepth === null
+        || requestedMaxDepth === ''
+        || (Number.isFinite(parsedMaxDepth) && parsedMaxDepth >= 15)
+      )
+      && toolResult.truncated !== true
+      && toolResult.hasMore !== true
+      && toolResult.autoDegraded !== true;
+    if (
+      !challenge
+      && activeGate
+      && authoritativeRootRead
+      && Number.isInteger(activeGate.challengeFrameId)
+    ) {
+      const frameInspection = await this._detectChallengeDialogBeforeMutation(tabId, {
+        includeStatus: true,
+        expectedFrameId: activeGate.challengeFrameId,
+      });
+      if (frameInspection.challenge?.label) {
+        challenge = detectChallengeDialog(
+          `dialog ${JSON.stringify(String(frameInspection.challenge.label).slice(0, 200))}`
+        );
+      } else if (!frameInspection.inspectionComplete) {
+        const guardedState = normalizeCaptchaGateState(activeGate);
+        this._captchaGateStates.set(tabId, guardedState);
+        toolResult.captchaGate = guardedState.publicGate;
+        return { gate: guardedState.publicGate, loopCheck: { kind: 'none' } };
+      }
+    }
+    const correlatedCaptchaCandidateIdentity =
+      activeGate?.publicGate?.candidateNotCorrelated === true
+        ? null
+        : activeGate?.captchaCandidateIdentity;
     let postSolveTokenState = null;
     if (
       ['verification_pending', 'manual_required', 'cleared'].includes(activeGate?.status)
-      && activeGate.captchaCandidateIdentity
+      && correlatedCaptchaCandidateIdentity
     ) {
       await inspectCaptchaFrames();
       if (!detectionFailed && detection && typeof detection === 'object') {
         postSolveTokenState = captchaPostSolveTokenState(
           detection,
-          activeGate.captchaCandidateIdentity,
+          correlatedCaptchaCandidateIdentity,
         );
       }
     }
-    const loopCheck = challenge
+    const directCaptchaEvidence = !!correlatedCaptchaCandidateIdentity
+      || activeGate?.publicGate?.languageNeutralFrameTrigger === true;
+    const loopCheck = challenge || (authoritativeRootRead && !directCaptchaEvidence)
       ? this._checkVerificationChallengeLoop(tabId, {
           pageUrl,
           dialogLabel: challenge?.normalizedLabel || '',
@@ -4493,14 +4534,9 @@ Rules: no prose intro, no conclusion, no "this screenshot shows...", no layout d
       return { gate: pendingGate, loopCheck };
     }
     if (!challenge) {
-      if (
-        activeGate?.status === 'manual_required'
-        && !activeGate.captchaCandidateIdentity
-        && activeGate.publicGate?.languageNeutralFrameTrigger !== true
-      ) {
+      if (activeGate?.status === 'manual_required' && !directCaptchaEvidence) {
         const manualInspection = await this._detectChallengeDialogBeforeMutation(tabId, {
           includeStatus: true,
-          allowGenericFailure: true,
           ...(Number.isInteger(activeGate.challengeFrameId)
             ? { expectedFrameId: activeGate.challengeFrameId }
             : {}),
@@ -4516,6 +4552,12 @@ Rules: no prose intro, no conclusion, no "this screenshot shows...", no layout d
           return { gate: clearedGate, loopCheck };
         }
       }
+      if (activeGate && authoritativeRootRead && activeGate.status !== 'cleared') {
+        const guardedState = normalizeCaptchaGateState(activeGate);
+        this._captchaGateStates.set(tabId, guardedState);
+        toolResult.captchaGate = guardedState.publicGate;
+        return { gate: guardedState.publicGate, loopCheck };
+      }
       if (activeGate?.status === 'cleared') {
         return { gate: null, loopCheck };
       }
@@ -4609,7 +4651,9 @@ Rules: no prose intro, no conclusion, no "this screenshot shows...", no layout d
       ...(detection?.selected && !selectedCorrelated ? { candidateNotCorrelated: true } : {}),
       ...(languageNeutralFrameTrigger ? { languageNeutralFrameTrigger: true } : {}),
     };
-    const captchaCandidateIdentity = captchaGateCandidateIdentity(detection?.selected);
+    const captchaCandidateIdentity = selectedCorrelated
+      ? captchaGateCandidateIdentity(detection?.selected)
+      : null;
     this._captchaGateStates.set(tabId, {
       key,
       status: publicGate.status,

src/chrome/src/agent/captcha-frame-runtime.js

@@ -602,7 +602,7 @@ function selectedReason(candidate, constraints) {
 
 // This function is serialized and executed in the web page. It must not
 // reference module-scope values.
-export function detectCaptchaCandidatesInPage(scope = null) {
+export function detectCaptchaCandidatesInPage(scope = null, matcherOptions = null) {
   const candidates = [];
   const pageWindow = scope?.window
     || (typeof window !== 'undefined' ? window : null);
@@ -695,10 +695,15 @@ export function detectCaptchaCandidatesInPage(scope = null) {
       return false;
     }
   };
-  const challengeDialogRe = /\b(?:(?:re|h|fun)?captcha|security verification|human verification|verify (?:that )?you(?:'|\u2019)re (?:a )?human|verify (?:that )?you are (?:a )?human|are you (?:a )?human|robot check|challenge verification)\b/i;
+  let challengeDialogRe = null;
+  try {
+    const source = String(matcherOptions?.challengeLabelPatternSource || '');
+    if (source) challengeDialogRe = new RegExp(source, 'i');
+  } catch (_) {}
   const challengeDialogs = Array.from(
     pageDocument.querySelectorAll('dialog, [role="dialog"], [role="alertdialog"]')
   ).filter((element) => {
+    if (!challengeDialogRe) return false;
     if (!visibleElement(element)) return false;
     let labelledBy = '';
     try {

src/chrome/src/agent/captcha-gate.js

@@ -1,16 +1,14 @@
 import { applyCaptchaFrameVisibility } from './captcha-frame-runtime.js';
 
-const CHALLENGE_DIALOG_RE = /\b(?:(?:re|h|fun)?captcha|security verification|human verification|verify (?:that )?you(?:'|\u2019)re (?:a )?human|verify (?:that )?you are (?:a )?human|are you (?:a )?human|robot check|challenge verification)\b/i;
-const CHALLENGE_FAILURE_RE = /\b(?:verification (?:failed|error|unsuccessful|expired|timed out)|could not verify|unable to verify)\b/i;
-const CHALLENGE_CONTEXT_RE = /\b(?:(?:re|h|fun)?captcha|human|robot|challenge)\b/i;
+export const CAPTCHA_CHALLENGE_LABEL_PATTERN_SOURCE = String.raw`\b(?:(?:re|h|fun)?captcha|security verification|human verification|verify (?:that )?you(?:'|\u2019)re (?:a )?human|verify (?:that )?you are (?:a )?human|are you (?:a )?human|robot check|challenge verification)\b`;
+const CHALLENGE_DIALOG_RE = new RegExp(CAPTCHA_CHALLENGE_LABEL_PATTERN_SOURCE, 'i');
 
-function matchesChallengeLabel(value, allowGenericFailure = false) {
-  const text = String(value || '');
-  return CHALLENGE_DIALOG_RE.test(text)
-    || (
-      CHALLENGE_FAILURE_RE.test(text)
-      && (allowGenericFailure || CHALLENGE_CONTEXT_RE.test(text))
-    );
+export function captchaChallengeMatcherOptions() {
+  return { challengeLabelPatternSource: CAPTCHA_CHALLENGE_LABEL_PATTERN_SOURCE };
+}
+
+function matchesChallengeLabel(value) {
+  return CHALLENGE_DIALOG_RE.test(String(value || ''));
 }
 
 function normalizeChallengeLabel(value) {
@@ -47,16 +45,15 @@ function parseSerializedTreeLabel(line) {
   return '';
 }
 
-export function detectChallengeDialog(pageContent, options = null) {
-  const allowGenericFailure = options?.allowGenericFailure === true;
+export function detectChallengeDialog(pageContent) {
   const lines = String(pageContent || '').split(/\r?\n/);
   for (let index = 0; index < lines.length; index += 1) {
     const line = lines[index];
     const dialogMatch = line.match(/^(\s*)(?:dialog|alertdialog)(?=\s|$)/i);
     if (!dialogMatch) continue;
     const dialogIndent = dialogMatch[1].length;
     const ownLabel = parseSerializedTreeLabel(line);
-    if (ownLabel && matchesChallengeLabel(ownLabel, allowGenericFailure)) {
+    if (ownLabel && matchesChallengeLabel(ownLabel)) {
       return {
         label: ownLabel,
         normalizedLabel: normalizeChallengeLabel(ownLabel),
@@ -68,7 +65,7 @@ export function detectChallengeDialog(pageContent, options = null) {
       const childIndent = childLine.match(/^\s*/)?.[0].length || 0;
       if (childIndent <= dialogIndent) break;
       const childLabel = parseSerializedTreeLabel(childLine);
-      if (!childLabel || !matchesChallengeLabel(childLabel, allowGenericFailure)) continue;
+      if (!childLabel || !matchesChallengeLabel(childLabel)) continue;
       return {
         label: childLabel,
         normalizedLabel: normalizeChallengeLabel(childLabel),
@@ -82,7 +79,6 @@ export function detectChallengeDialog(pageContent, options = null) {
 // model-authored mutations. Keep this function self-contained.
 export function detectChallengeDialogInPage(options = null) {
   const includeFrameContext = options?.includeFrameContext === true;
-  const allowGenericFailure = options?.allowGenericFailure === true;
   const pageWindow = typeof window !== 'undefined' ? window : null;
   const pageLocation = pageWindow?.location
     || (typeof location !== 'undefined' ? location : null);
@@ -96,16 +92,13 @@ export function detectChallengeDialogInPage(options = null) {
       ? { challenge: null, frameContext: { frameUrl, frameName, childFrames: [] } }
       : null;
   }
-  const challengeRe = /\b(?:(?:re|h|fun)?captcha|security verification|human verification|verify (?:that )?you(?:'|\u2019)re (?:a )?human|verify (?:that )?you are (?:a )?human|are you (?:a )?human|robot check|challenge verification)\b/i;
-  const challengeFailureRe = /\b(?:verification (?:failed|error|unsuccessful|expired|timed out)|could not verify|unable to verify)\b/i;
-  const challengeContextRe = /\b(?:(?:re|h|fun)?captcha|human|robot|challenge)\b/i;
+  let challengeRe = null;
+  try {
+    const source = String(options?.challengeLabelPatternSource || '');
+    if (source) challengeRe = new RegExp(source, 'i');
+  } catch {}
   const matchesChallenge = value => {
-    const text = String(value || '');
-    return challengeRe.test(text)
-      || (
-        challengeFailureRe.test(text)
-        && (allowGenericFailure || challengeContextRe.test(text))
-      );
+    return challengeRe?.test(String(value || '')) === true;
   };
   const visible = (element) => {
     try {

src/chrome/src/agent/captcha-solver.js

@@ -22,7 +22,7 @@ import {
   normalizeCaptchaType,
   selectCaptchaCandidate,
 } from './captcha-frame-runtime.js';
-import { buildCaptchaDiagnostics } from './captcha-gate.js';
+import { buildCaptchaDiagnostics, captchaChallengeMatcherOptions } from './captcha-gate.js';
 
 export { captchaTypesMatch, captchaWebsiteUrl, normalizeCaptchaType, selectCaptchaCandidate };
 
@@ -258,6 +258,7 @@ export async function detectCaptcha(tabId, constraints = {}) {
     chrome.scripting.executeScript({
       target: { tabId, allFrames: true },
       func: detectCaptchaCandidatesInPage,
+      args: [null, captchaChallengeMatcherOptions()],
     }),
     frameTreePromise,
   ]);