pi-agent-browser-native 0.5.0 → 0.6.6

package/dist/extensions/agent-browser/lib/process.js

@@ -4,12 +4,11 @@ import { lstat, mkdir, readdir } from "node:fs/promises";
 import { dirname, isAbsolute, join } from "node:path";
 import { env as processEnv, platform as processPlatform } from "node:process";
 import { parseArgvDescriptor } from "./argv-descriptor.js";
-import { needsManagedSession } from "./command-policy.js";
 import { isKnownCommandToken } from "./command-taxonomy.js";
 import { extractExplicitSessionName, getFlagName, GLOBAL_BOOLEAN_FLAGS_WITH_OPTIONAL_VALUES, GLOBAL_VALUE_FLAGS, optionalGlobalValueFlagConsumesNext, resolveAgentBrowserNamespace, } from "./argv-grammar.js";
-import { canonicalizeOwnedManagedSessionCloseArgs, commitManagedSessionRestoreSuppression, getManagedSessionRestoreConfigEnv, getManagedSessionRestoreEnv, getOwnedManagedSessionRestoreKey, getManagedSessionRestoreProtectedEnv, getOwnedManagedSessionCompatibilityEnv, getOwnedManagedSessionNamespaceEnv, isOwnedManagedSessionTarget, shouldOmitOwnedManagedSessionRestoreEnv, validateManagedSessionRestoreContextForSpawn, } from "./managed-session-restore.js";
-import { getManagedSessionStateAccessValidationError, getManagedSessionTargetAccessValidationError, invocationMayNavigateToLocalFile, } from "./managed-session-state-policy.js";
-import { getImplicitSessionIdleTimeoutMs, isPlainTextInspectionArgs } from "./runtime.js";
+import { commitManagedSessionRestoreSuppression, getManagedSessionRestoreEnv, getManagedSessionRestoreProtectedEnv, getOwnedManagedSessionCompatibilityEnv, getOwnedManagedSessionNamespaceEnv, isOwnedManagedSessionTarget, validateManagedSessionRestoreContextForSpawn, } from "./managed-session-restore.js";
+import { getPageTargetValidationError, } from "./page-target-validation.js";
+import { getImplicitSessionIdleTimeoutMs } from "./runtime.js";
 import { getAgentBrowserProcessEnvironment } from "./process-environment.js";
 import { openSecureTempFile, writeSecureTempChunk } from "./temp.js";
 const MAX_BUFFERED_STDOUT_BYTES = 512 * 1_024;
@@ -17,7 +16,6 @@ const MAX_BUFFERED_STDERR_CHARS = 32_000;
 const MAX_BUFFERED_STDOUT_TAIL_CHARS = 32_000;
 const PROCESS_STDOUT_SPILL_FILE_PREFIX = "process-stdout";
 const AGENT_BROWSER_SOCKET_DIR_ENV = "AGENT_BROWSER_SOCKET_DIR";
-const AGENT_BROWSER_ARGS_ENV = "AGENT_BROWSER_ARGS";
 const AGENT_BROWSER_DEFAULT_TIMEOUT_ENV = "AGENT_BROWSER_DEFAULT_TIMEOUT";
 const AGENT_BROWSER_IDLE_TIMEOUT_ENV = "AGENT_BROWSER_IDLE_TIMEOUT_MS";
 const PI_AGENT_BROWSER_PROCESS_TIMEOUT_ENV = "PI_AGENT_BROWSER_PROCESS_TIMEOUT_MS";
@@ -29,12 +27,11 @@ const DEFAULT_AGENT_BROWSER_PROCESS_TIMEOUT_MS = 35_000;
 /** Grace period after `exit` before resolving when `close` is delayed by inherited stdio handles. */
 const EXIT_STDIO_GRACE_MS = 100;
 const WINDOWS_AGENT_BROWSER_MISSING_MARKER = "PI_AGENT_BROWSER_COMMAND_NOT_FOUND:agent-browser.cmd";
-const UNTRUSTED_LOCAL_FILE_SESSION_MESSAGE = "Local file navigation requires a wrapper-managed local browser because caller-owned or attached sessions can retain unsafe file-access launch flags. Omit the explicit session or attachment and retry with sessionMode fresh.";
 const attachedBrowserSessionContext = new AsyncLocalStorage();
 const WINDOWS_COMMANDS_WITH_ADJACENT_SUBCOMMAND = new Set([
     "auth", "clipboard", "cookies", "dashboard", "device", "dialog", "diff", "find", "get", "is", "keyboard",
     "mouse", "network", "plugin", "profiler", "react", "record", "session", "set", "skills", "state", "storage",
-    "stream", "tab", "trace", "window",
+    "stream", "tab", "trace", "webmcp", "window",
 ]);
 export function withAttachedBrowserSessionContext(preserve, run) {
     return attachedBrowserSessionContext.run(preserve || attachedBrowserSessionContext.getStore() === true, run);
@@ -111,27 +108,10 @@ export function reorderWindowsLeadingGlobalArgs(args) {
     }
     return args;
 }
-export function pinAgentBrowserFileAccessDisabled(args, wrapperCompatibilityUserAgent, preserveAttachedBrowserSession = false) {
-    const filtered = [];
-    for (let index = 0; index < args.length; index += 1) {
-        const token = args[index];
-        if (token.startsWith("--allow-file-access="))
-            continue;
-        if (token === "--allow-file-access") {
-            if (["false", "true"].includes(args[index + 1] ?? ""))
-                index += 1;
-            continue;
-        }
-        filtered.push(token);
-    }
-    // These are launch-only controls. Sending them on an attached-session follow-up makes upstream replace the CDP connection with a local browser.
-    if (preserveAttachedBrowserSession)
-        return filtered;
-    // Upstream's flag overrides only the active CDP target; the Chrome arg covers new tabs. Its --args parser splits commas/newlines.
-    const browserArgs = wrapperCompatibilityUserAgent
-        ? ["--args", `--user-agent=${wrapperCompatibilityUserAgent.replaceAll(/[\r\n,]/g, "")}`]
-        : [];
-    return [...browserArgs, "--allow-file-access", "false", ...filtered];
+export function prepareAgentBrowserSpawnArgs(args, wrapperCompatibilityUserAgent, preserveAttachedBrowserSession = false) {
+    if (preserveAttachedBrowserSession || !wrapperCompatibilityUserAgent)
+        return args;
+    return ["--args", `--user-agent=${wrapperCompatibilityUserAgent.replaceAll(/[\r\n,]/g, "")}`, ...args];
 }
 export function buildAgentBrowserSpawnCommand(args, platform = processPlatform) {
     if (platform !== "win32") {
@@ -384,54 +364,27 @@ export function buildAgentBrowserProcessEnv(baseEnv = processEnv, overrides = un
     clampUpstreamDefaultTimeout(childEnv);
     return childEnv;
 }
-function getManagedPreSpawnPolicyError(options, effectiveEnv, allowManagedSessionTarget = false, currentPageUrl, pageUrlUnknown = false, trustedFirstBatchTabSelection = false, trustedPinnedEmptyConfig = false) {
-    const policyEnv = effectiveEnv ?? { ...(options.parentEnv ?? processEnv), ...options.env };
-    const managedSessionTargetError = getManagedSessionTargetAccessValidationError(options.args, allowManagedSessionTarget || options.ownedManagedSession === true || isOwnedManagedSessionTarget(options.args), policyEnv);
-    if (managedSessionTargetError)
-        return managedSessionTargetError;
+function getManagedPreSpawnPolicyError(options, currentPageUrl, pageUrlUnknown = false, trustedFirstBatchTabSelection = false) {
     if (!validateManagedSessionRestoreContextForSpawn(options)) {
         return "Managed session restore policy, storage, or checkout identity changed after planning; refusing to start agent-browser.";
     }
-    return getManagedSessionStateAccessValidationError({
+    return getPageTargetValidationError({
         args: options.args,
         currentPageUrl,
-        cwd: options.cwd,
-        env: effectiveEnv ?? options.env,
-        managedSessionRestoreKey: getOwnedManagedSessionRestoreKey(),
         pageUrlUnknown,
-        parentEnv: effectiveEnv ? {} : options.parentEnv ?? processEnv,
         stdin: options.stdin,
         trustedFirstBatchTabSelection,
-        trustedPinnedEmptyConfig,
     });
 }
 export async function runAgentBrowserProcess(options) {
-    const { allowManagedSessionTarget, cwd, env, managedSessionRestoreState, managedStateCurrentPageUrl, managedStatePageUrlUnknown, signal, stdin, trustedFirstBatchTabSelection } = options;
+    const { cwd, env, managedSessionRestoreState, managedStateCurrentPageUrl, managedStatePageUrlUnknown, signal, stdin, trustedFirstBatchTabSelection } = options;
     const preserveAttachedBrowserSession = options.preserveAttachedBrowserSession === true || attachedBrowserSessionContext.getStore() === true;
     const ownedManagedSession = options.ownedManagedSession === true || isOwnedManagedSessionTarget(options.args);
-    const args = canonicalizeOwnedManagedSessionCloseArgs({
-        args: options.args,
-        cwd,
-        env,
-        ownedManagedSession,
-        restoreState: managedSessionRestoreState,
-        stdin,
-    });
+    const args = options.args;
     const timeoutMs = options.timeoutMs ?? getAgentBrowserProcessTimeoutMs();
     if (signal?.aborted) {
         return { aborted: true, agentBrowserStarted: false, exitCode: 1, stderr: "", stdout: "", timedOut: false };
     }
-    if (invocationMayNavigateToLocalFile(args, stdin) && (!ownedManagedSession || preserveAttachedBrowserSession)) {
-        return {
-            aborted: false,
-            agentBrowserStarted: false,
-            exitCode: 1,
-            spawnError: new Error(UNTRUSTED_LOCAL_FILE_SESSION_MESSAGE),
-            stderr: "",
-            stdout: "",
-            timedOut: false,
-        };
-    }
     const parentEnv = getAgentBrowserProcessEnvironment();
     const managedSessionRestoreOptions = {
         args,
@@ -442,7 +395,7 @@ export async function runAgentBrowserProcess(options) {
         restoreState: managedSessionRestoreState,
         stdin,
     };
-    const planningPolicyError = getManagedPreSpawnPolicyError(managedSessionRestoreOptions, undefined, allowManagedSessionTarget, managedStateCurrentPageUrl, managedStatePageUrlUnknown, trustedFirstBatchTabSelection);
+    const planningPolicyError = getManagedPreSpawnPolicyError(managedSessionRestoreOptions, managedStateCurrentPageUrl, managedStatePageUrlUnknown, trustedFirstBatchTabSelection);
     if (planningPolicyError) {
         return {
             aborted: false,
@@ -455,37 +408,18 @@ export async function runAgentBrowserProcess(options) {
         };
     }
     const managedSessionRestoreEnv = getManagedSessionRestoreEnv(managedSessionRestoreOptions);
-    const ownedManagedSessionClose = shouldOmitOwnedManagedSessionRestoreEnv(managedSessionRestoreOptions);
-    const browserConfigPinRequired = !isPlainTextInspectionArgs(args) && needsManagedSession(parseArgvDescriptor(args));
-    const managedSessionRestoreConfigEnv = await getManagedSessionRestoreConfigEnv(managedSessionRestoreEnv, ownedManagedSessionClose || browserConfigPinRequired);
-    if (managedSessionRestoreConfigEnv === undefined) {
-        return {
-            aborted: false,
-            agentBrowserStarted: false,
-            exitCode: 1,
-            spawnError: new Error("Browser-backed agent-browser commands require a protected empty config, but secure temp storage was unavailable."),
-            stderr: "",
-            stdout: "",
-            timedOut: false,
-        };
-    }
     const ownedManagedSessionCompatibilityEnv = getOwnedManagedSessionCompatibilityEnv(managedSessionRestoreOptions);
     const processOverrides = {
         [AGENT_BROWSER_IDLE_TIMEOUT_ENV]: String(getImplicitSessionIdleTimeoutMs()),
         ...managedSessionRestoreEnv,
         ...env,
-        ...managedSessionRestoreConfigEnv,
         ...getManagedSessionRestoreProtectedEnv(managedSessionRestoreOptions, managedSessionRestoreEnv),
         ...getOwnedManagedSessionNamespaceEnv(managedSessionRestoreOptions),
         ...getWindowsExplicitDefaultNamespaceEnv(args, parentEnv.AGENT_BROWSER_NAMESPACE),
         ...ownedManagedSessionCompatibilityEnv,
-        AGENT_BROWSER_ALLOW_FILE_ACCESS: undefined,
-        [AGENT_BROWSER_ARGS_ENV]: undefined,
     };
     const explicitSocketDir = processOverrides[AGENT_BROWSER_SOCKET_DIR_ENV];
     let effectiveEnv = explicitSocketDir === undefined ? { ...processOverrides, [AGENT_BROWSER_SOCKET_DIR_ENV]: undefined } : processOverrides;
-    if (ownedManagedSessionClose)
-        effectiveEnv = { ...effectiveEnv, AGENT_BROWSER_RESTORE: undefined };
     const requestedSocketDir = explicitSocketDir ?? parentEnv[PI_AGENT_BROWSER_SOCKET_DIR_ENV] ?? getAgentBrowserSocketDir();
     if (requestedSocketDir !== undefined) {
         const socketDirError = requestedSocketDir.length > 0
@@ -522,6 +456,7 @@ export async function runAgentBrowserProcess(options) {
         let stdoutTail = "";
         let stdoutSpillHandle;
         let stdoutSpillPath;
+        let stdoutSpillPending = false;
         let pendingStdoutWrite = Promise.resolve();
         let stdoutSpillError;
         let killTimer;
@@ -533,11 +468,12 @@ export async function runAgentBrowserProcess(options) {
             stdoutTail = appendTail(stdoutTail, buffer.toString("utf8"), MAX_BUFFERED_STDOUT_TAIL_CHARS);
             if (stdoutSpillError)
                 return;
-            if (!stdoutSpillPath && stdoutBufferedBytes + buffer.length <= MAX_BUFFERED_STDOUT_BYTES) {
+            if (!stdoutSpillPending && !stdoutSpillPath && stdoutBufferedBytes + buffer.length <= MAX_BUFFERED_STDOUT_BYTES) {
                 stdoutBuffers.push(buffer);
                 stdoutBufferedBytes += buffer.length;
                 return;
             }
+            stdoutSpillPending = true;
             pendingStdoutWrite = pendingStdoutWrite
                 .then(async () => {
                 if (stdoutSpillError)
@@ -612,12 +548,12 @@ export async function runAgentBrowserProcess(options) {
             });
         };
         const childEnv = buildAgentBrowserProcessEnv(parentEnv, effectiveEnv);
-        const spawnPolicyError = getManagedPreSpawnPolicyError(managedSessionRestoreOptions, childEnv, allowManagedSessionTarget, managedStateCurrentPageUrl, managedStatePageUrlUnknown, trustedFirstBatchTabSelection, managedSessionRestoreConfigEnv.AGENT_BROWSER_CONFIG !== undefined);
+        const spawnPolicyError = getManagedPreSpawnPolicyError(managedSessionRestoreOptions, managedStateCurrentPageUrl, managedStatePageUrlUnknown, trustedFirstBatchTabSelection);
         if (spawnPolicyError) {
             resolve({ aborted: false, agentBrowserStarted: false, exitCode: 1, spawnError: new Error(spawnPolicyError), stderr: "", stdout: "", timedOut: false });
             return;
         }
-        const spawnCommand = buildAgentBrowserSpawnCommand(pinAgentBrowserFileAccessDisabled(args, ownedManagedSessionCompatibilityEnv.AGENT_BROWSER_USER_AGENT, preserveAttachedBrowserSession));
+        const spawnCommand = buildAgentBrowserSpawnCommand(prepareAgentBrowserSpawnArgs(args, ownedManagedSessionCompatibilityEnv.AGENT_BROWSER_USER_AGENT, preserveAttachedBrowserSession));
         const child = spawn(spawnCommand.command, spawnCommand.args, {
             cwd,
             env: childEnv,

package/dist/extensions/agent-browser/lib/recording-reservations.js

@@ -1,9 +1,6 @@
-import { extractUpstreamCommandTokens } from "./argv-descriptor.js";
-import { getAgentBrowserSessionIdentityKey, isAgentBrowserSessionIdentityKeyInNamespace } from "./argv-grammar.js";
-import { batchHasSuccessfulCloseAll, getSuccessfulBatchCloseLifecycle } from "./batch-lifecycle.js";
-import { isCloseAllCommand, isCloseCommand } from "./command-taxonomy.js";
+import { getAgentBrowserSessionIdentityKey } from "./argv-grammar.js";
 import { isRecord } from "./parsing.js";
-import { isPendingRecordingArtifact, isPendingRecordingCommand, isSessionArtifactManifest } from "./results/artifact-manifest.js";
+import { isPendingRecordingArtifact } from "./results/artifact-manifest.js";
 export const RECORDING_RESERVATION_ENTRY_TYPE = "agent-browser-recording-reservation";
 function getReservationKey(reservation) {
     return getAgentBrowserSessionIdentityKey(reservation.sessionName, reservation.namespace);
@@ -100,83 +97,19 @@ export function restoreRecordingReservationStateFromBranch(branch) {
     const reservations = new Map();
     const terminal = new Map();
     for (const entry of branch) {
-        if (!isRecord(entry))
+        if (!isRecord(entry) || entry.type !== "custom" || entry.customType !== RECORDING_RESERVATION_ENTRY_TYPE)
             continue;
-        if (entry.type === "custom" && entry.customType === RECORDING_RESERVATION_ENTRY_TYPE) {
-            const transition = parseReservationTransition(entry.data);
-            if (!transition)
-                continue;
-            const key = getReservationKey(transition.reservation);
-            if (transition.state === "active") {
-                reservations.set(key, transition.reservation);
-                terminal.delete(key);
-            }
-            else {
-                reservations.delete(key);
-                terminal.set(key, transition.reservation);
-            }
+        const transition = parseReservationTransition(entry.data);
+        if (!transition)
             continue;
+        const key = getReservationKey(transition.reservation);
+        if (transition.state === "active") {
+            reservations.set(key, transition.reservation);
+            terminal.delete(key);
         }
-        if (entry.type !== "message")
-            continue;
-        const message = isRecord(entry.message) ? entry.message : undefined;
-        const details = message?.toolName === "agent_browser" && isRecord(message.details) ? message.details : undefined;
-        if (!details)
-            continue;
-        if (isSessionArtifactManifest(details.artifactManifest)) {
-            const artifacts = [...details.artifactManifest.entries]
-                .sort((left, right) => left.createdAtMs - right.createdAtMs
-                || Number(isPendingRecordingCommand(left.command, left.subcommand, left.kind)) - Number(isPendingRecordingCommand(right.command, right.subcommand, right.kind))
-                || left.path.localeCompare(right.path))
-                .filter((manifestEntry) => manifestEntry.command === "record" && manifestEntry.kind === "video" && manifestEntry.session)
-                .map((manifestEntry) => ({
-                absolutePath: manifestEntry.absolutePath ?? manifestEntry.path,
-                command: manifestEntry.command,
-                cwd: manifestEntry.cwd,
-                kind: "video",
-                namespace: manifestEntry.namespace,
-                path: manifestEntry.path,
-                session: manifestEntry.session,
-                status: isPendingRecordingCommand(manifestEntry.command, manifestEntry.subcommand, "video") ? "pending" : "saved",
-                subcommand: manifestEntry.subcommand,
-            }));
-            for (const artifact of artifacts) {
-                const reservation = getArtifactReservation(artifact);
-                if (!reservation)
-                    continue;
-                const key = getReservationKey(reservation);
-                applyRecordingArtifactsToReservations(reservations, [artifact]);
-                if (isPendingRecordingArtifact(artifact))
-                    terminal.delete(key);
-                else
-                    terminal.set(key, reservation);
-            }
-        }
-        const messageSucceeded = typeof message?.isError === "boolean"
-            ? !message.isError
-            : typeof details.exitCode !== "number" || details.exitCode === 0;
-        const args = Array.isArray(details.args) && details.args.every((arg) => typeof arg === "string") ? details.args : [];
-        const namespace = typeof details.namespace === "string" ? details.namespace : undefined;
-        const closeAllApplied = details.closeAllApplied === true
-            || (messageSucceeded && isCloseAllCommand(extractUpstreamCommandTokens(args)))
-            || batchHasSuccessfulCloseAll(details.batchSteps);
-        if (closeAllApplied) {
-            for (const [key, reservation] of reservations) {
-                if (!isAgentBrowserSessionIdentityKeyInNamespace(key, namespace))
-                    continue;
-                terminal.set(key, reservation);
-                reservations.delete(key);
-            }
-            continue;
-        }
-        const directCloseSucceeded = messageSucceeded && isCloseCommand(typeof details.command === "string" ? details.command : undefined);
-        const batchRecordingClosed = getSuccessfulBatchCloseLifecycle(details.batchSteps)?.recordingClosedAfterBatch === true;
-        if ((directCloseSucceeded || batchRecordingClosed) && typeof details.sessionName === "string") {
-            const key = getAgentBrowserSessionIdentityKey(details.sessionName, namespace);
-            const reservation = reservations.get(key);
-            if (reservation)
-                terminal.set(key, reservation);
+        else {
             reservations.delete(key);
+            terminal.set(key, transition.reservation);
         }
     }
     return { active: reservations, terminal };

package/dist/extensions/agent-browser/lib/results/action-recommendations.js

@@ -106,7 +106,7 @@ export function buildAgentBrowserNextActions(options) {
                 reason: "Inspect the opened page before choosing interactive refs.",
             }));
         }
-        else if (isPageMutationCommand(options.command)) {
+        else if (isPageMutationCommand(options.command, options.subcommand)) {
             actions.push(buildNextToolAction({
                 args: ["snapshot", "-i"],
                 id: "inspect-after-mutation",

package/dist/extensions/agent-browser/lib/results/presentation/batch.js

@@ -3,12 +3,13 @@ import { isRecord } from "../../parsing.js";
 import { getAgentBrowserSessionIdentityKey } from "../../argv-grammar.js";
 import { extractUpstreamCommandTokens, parseCommandInfo, redactInvocationArgs, redactSensitiveText, redactSensitiveValue } from "../../runtime.js";
 import { buildAgentBrowserNextActions } from "../action-recommendations.js";
+import { buildPendingWebMcpNextActions } from "../recovery-next-actions.js";
 import { formatSessionArtifactRetentionSummary, isPendingRecordingArtifact } from "../artifact-manifest.js";
 import { classifyAgentBrowserFailureCategory } from "../categories.js";
 import { detectConfirmationRequired } from "../confirmation.js";
 import { applyNetworkRouteRecords, buildNetworkRouteDiagnostics } from "../network-routes.js";
 import { appendUniqueAgentBrowserNextActions, applyNamespaceToNextActions, withOptionalSessionArgs } from "../next-actions.js";
-import { stringifyModelFacing } from "./common.js";
+import { extractAgentBrowserLifecycle, stringifyModelFacing } from "./common.js";
 import { buildArtifactVerificationSummary, classifyPresentationSuccessCategory, manifestHasNewNoticeWorthyEntries } from "./artifacts.js";
 import { formatBatchStepCommand, getPresentationImages, getPresentationPaths, getPresentationText, isStringArray } from "./content.js";
 import { buildPageChangeSummary } from "./navigation.js";
@@ -94,6 +95,11 @@ function redactExactValues(value, sensitiveValues) {
         return value;
     return redactSensitiveValue(Object.fromEntries(Object.entries(value).map(([key, entryValue]) => [key, redactExactValues(entryValue, sensitiveValues)])));
 }
+export function redactBatchStepErrorData(command, value) {
+    return command?.[0] === "clipboard"
+        ? redactSensitiveValue(redactClipboardPermissionErrorValue({ command: "clipboard", subcommand: command[1] }, value, getClipboardWritePayloadCandidates(command)))
+        : redactExactValues(value, getStatefulCommandSensitiveValues(command));
+}
 function getTypedTextLength(command) {
     return command?.[0] === "keyboard" && command[1] === "type" && typeof command[2] === "string"
         ? Array.from(command[2]).length
@@ -187,11 +193,9 @@ async function buildBatchStepPresentation(options) {
     const command = isStringArray(item.command) ? item.command : undefined;
     const redactedCommand = command ? redactInvocationArgs(command) : undefined;
     const commandText = formatBatchStepCommand(hasModelFacingArgRedaction(redactedCommand) ? redactedCommand : command, index);
-    const lifecycle = extractBatchStepLifecycle(item.result);
+    const lifecycle = extractAgentBrowserLifecycle(item.result);
     if (item.success === false) {
-        const redactedErrorData = command?.[0] === "clipboard"
-            ? redactSensitiveValue(redactClipboardPermissionErrorValue({ command: "clipboard", subcommand: command[1] }, item.error, getClipboardWritePayloadCandidates(command)))
-            : redactExactValues(item.error, getStatefulCommandSensitiveValues(command));
+        const redactedErrorData = redactBatchStepErrorData(command, item.error);
         const errorText = formatBatchStepError(redactedErrorData);
         const failureCategory = classifyAgentBrowserFailureCategory({
             args: command,
@@ -206,6 +210,7 @@ async function buildBatchStepPresentation(options) {
             failureCategory,
             resultCategory: "failure",
             sessionName,
+            subcommand: command?.[1],
         })), namespace);
         const presentation = {
             content: [{ type: "text", text: errorText }],
@@ -262,7 +267,12 @@ async function buildBatchStepPresentation(options) {
     });
     const text = getPresentationText(presentation) || presentation.summary;
     const stepSucceeded = presentation.resultCategory !== "failure";
-    const nextActions = applyNamespaceToNextActions(presentation.nextActions ?? buildAgentBrowserNextActions({
+    const pendingWebMcpMutation = stepSucceeded
+        && commandInfo.command === "webmcp"
+        && ["invoke", "result"].includes(commandInfo.subcommand ?? "")
+        && isRecord(presentation.data)
+        && presentation.data.status === "pending";
+    const nextActions = applyNamespaceToNextActions(pendingWebMcpMutation ? buildPendingWebMcpNextActions(sessionName) : presentation.nextActions ?? buildAgentBrowserNextActions({
         artifacts: presentation.artifacts,
         args: command,
         command: command?.[0],
@@ -270,6 +280,7 @@ async function buildBatchStepPresentation(options) {
         resultCategory: stepSucceeded ? "success" : "failure",
         savedFilePath: presentation.savedFilePath,
         sessionName,
+        subcommand: command?.[1],
         successCategory: presentation.successCategory,
     }), namespace);
     const pageChangeSummary = buildPageChangeSummary({
@@ -308,12 +319,6 @@ async function buildBatchStepPresentation(options) {
         presentation,
     };
 }
-function extractBatchStepLifecycle(result) {
-    if (!isRecord(result) || !isRecord(result.lifecycle) || !isRecord(result.lifecycle.effectiveLaunch))
-        return undefined;
-    const browserLaunched = result.lifecycle.effectiveLaunch.browserLaunched;
-    return typeof browserLaunched === "boolean" ? { effectiveLaunch: { browserLaunched } } : undefined;
-}
 function abandonedRecordingArtifact(artifact) {
     const { recordingState: _recordingState, willExistOnStop: _willExistOnStop, ...terminal } = artifact;
     return { ...terminal, exists: false, status: "missing", subcommand: "close-abandoned" };
@@ -394,6 +399,10 @@ export async function buildBatchPresentation(options) {
     const redactedBatchData = steps.map(({ details }) => (details.success
         ? { command: details.command, result: details.data, success: true }
         : { command: details.command, error: details.text, success: false }));
+    const unverifiedMutationCount = steps.filter((step) => step.details.pageChangeSummary?.changeType === "mutation" && step.details.pageChangeSummary.observed === false).length;
+    const mutationEvidenceText = unverifiedMutationCount > 0
+        ? `Mutation evidence: ${unverifiedMutationCount} action result${unverifiedMutationCount === 1 ? " proves" : "s prove"} dispatch only, not application state change. Use explicit later assertions or external receipts as postconditions; fixed waits are not postconditions.`
+        : undefined;
     const stepText = formatBatchStepsText(steps);
     const batchSummary = batchFailure === undefined
         ? summary
@@ -405,7 +414,7 @@ export async function buildBatchPresentation(options) {
             `First failing step: ${batchFailure.failedStep.index + 1} — ${batchFailure.failedStep.commandText}`,
             batchFailure.failureCount > 1 ? `${batchFailure.failureCount} steps failed. See the per-step results below.` : "See the per-step results below.",
         ].join("\n");
-    const text = failureHeader ? `${failureHeader}\n\n${stepText}` : stepText;
+    const text = [failureHeader, mutationEvidenceText, stepText].filter((line) => line !== undefined).join("\n\n");
     const artifactRetentionSummary = currentArtifactManifest ? formatSessionArtifactRetentionSummary(currentArtifactManifest) : undefined;
     const contentText = artifactRetentionSummary && manifestHasNewNoticeWorthyEntries(options.artifactManifest, currentArtifactManifest)
         ? `${text}\n\n${artifactRetentionSummary}`
@@ -415,6 +424,7 @@ export async function buildBatchPresentation(options) {
         ? appendUniqueAgentBrowserNextActions([...(batchFailure.failedStep.nextActions ?? [])], artifactLifecycleActions)
         : artifactLifecycleActions;
     const changedSteps = steps.map((step) => step.details).filter((details) => details.pageChangeSummary !== undefined);
+    const observedChangeCount = changedSteps.filter((details) => details.pageChangeSummary?.observed === true).length;
     const pageChangeSummary = artifacts.length > 0
         ? buildPageChangeSummary({
             artifacts,
@@ -428,7 +438,10 @@ export async function buildBatchPresentation(options) {
                 changeType: "mutation",
                 command: "batch",
                 nextActionIds: nextActions?.map((action) => action.id),
-                summary: `batch → mutation → ${changedSteps.length} changed step${changedSteps.length === 1 ? "" : "s"}`,
+                observed: observedChangeCount > 0,
+                summary: observedChangeCount > 0
+                    ? `batch → ${observedChangeCount} observed change${observedChangeCount === 1 ? "" : "s"}${unverifiedMutationCount > 0 ? `; ${unverifiedMutationCount} dispatched action${unverifiedMutationCount === 1 ? "" : "s"} unverified` : ""}`
+                    : `batch → ${unverifiedMutationCount} action${unverifiedMutationCount === 1 ? "" : "s"} dispatched → application change unverified`,
             }
             : undefined;
     return {

package/dist/extensions/agent-browser/lib/results/presentation/common.js

@@ -1,4 +1,4 @@
-import { containsManagedSessionRestoreKey } from "../../managed-session-capabilities.js";
+import { isRecord } from "../../parsing.js";
 import { redactSensitiveText, redactSensitiveValue } from "../../runtime.js";
 import { stringifyUnknown, truncateText } from "../text.js";
 const UNTITLED_PAGE_SUMMARY = "(untitled page)";
@@ -24,7 +24,7 @@ export function redactModelFacingText(text) {
     return redactSensitiveText(text);
 }
 export function redactModelFacingTextIfSensitive(text) {
-    return containsManagedSessionRestoreKey(text) || /(?:@|\b(?:access[_-]?key|api[_-]?key|auth|authorization|basic|bearer|connection[_-]?string|cookie|database[_-]?url|db[_-]?url|mongo(?:db)?[_-]?uri|pass(?:word)?|private[_-]?key|redis[_-]?url|secret|session[_-]?id|token)\b)/i.test(text)
+    return /(?:@|\b(?:access[_-]?key|api[_-]?key|auth|authorization|basic|bearer|connection[_-]?string|cookie|database[_-]?url|db[_-]?url|mongo(?:db)?[_-]?uri|pass(?:word)?|private[_-]?key|redis[_-]?url|secret|session[_-]?id|token)\b)/i.test(text)
         ? redactModelFacingText(text)
         : text;
 }
@@ -37,6 +37,24 @@ export function getStringField(data, key) {
 }
 // `lifecycle` is upstream launch/reuse bookkeeping, never page content, so it must not be the
 // answer an agent reads when a command has no dedicated presenter.
+export function extractAgentBrowserLifecycle(result) {
+    if (Array.isArray(result)) {
+        let latest;
+        for (const row of result) {
+            const nested = isRecord(row) ? row.result ?? row.data ?? row : row;
+            latest = extractAgentBrowserLifecycle(nested) ?? latest;
+        }
+        return latest;
+    }
+    if (!isRecord(result))
+        return undefined;
+    if (isRecord(result.lifecycle) && isRecord(result.lifecycle.effectiveLaunch)) {
+        const browserLaunched = result.lifecycle.effectiveLaunch.browserLaunched;
+        if (typeof browserLaunched === "boolean")
+            return { effectiveLaunch: { browserLaunched } };
+    }
+    return extractAgentBrowserLifecycle(result.result ?? result.data);
+}
 export function omitUpstreamLifecycle(data) {
     const { lifecycle: _lifecycle, ...rest } = data;
     return rest;

package/dist/extensions/agent-browser/lib/results/presentation/diagnostics.js

@@ -4,7 +4,6 @@ import { classifyNetworkRequestFailure, isApiLikeNetworkRequest, isNetworkArtifa
 import { withOptionalSessionArgs } from "../next-actions.js";
 import { stringifyUnknown, truncateText } from "../text.js";
 import { firstLine, formatCount, getArrayField, getStringField, parseJsonPreviewString, redactModelFacingText, redactModelFacingTextIfSensitive, stringifyModelFacing, } from "./common.js";
-import { filterCallerOwnedSessionListItems, filterCallerOwnedStateListItems, filterManagedSessionListRows, filterManagedStateListRows, } from "./managed-list-filter.js";
 const DIAGNOSTIC_REQUEST_PREVIEW_LIMIT = 40;
 const DIAGNOSTIC_LOG_PREVIEW_LIMIT = 80;
 const NETWORK_BODY_PREVIEW_MAX_CHARS = 280;
@@ -112,7 +111,7 @@ export function formatDiagnosticSummary(commandInfo, data) {
     if (commandInfo.command === "session") {
         const sessions = getArrayField(data, "sessions");
         if (sessions)
-            return `Sessions: ${filterCallerOwnedSessionListItems(sessions).length}`;
+            return `Sessions: ${sessions.length}`;
         const session = getStringField(data, "session");
         if (session)
             return `Session: ${session}`;
@@ -170,10 +169,8 @@ export function formatDiagnosticSummary(commandInfo, data) {
     }
     if (commandInfo.command === "state") {
         const states = getArrayField(data, "states") ?? getArrayField(data, "files");
-        if (states) {
-            const visibleStates = commandInfo.subcommand === "list" ? filterCallerOwnedStateListItems(states) : states;
-            return `States: ${visibleStates.length}`;
-        }
+        if (states)
+            return `States: ${states.length}`;
         if (commandInfo.subcommand === "load")
             return undefined;
         const stateName = getStringField(data, "name") ?? getStringField(data, "file") ?? getStringField(data, "filename") ?? getStringField(data, "path") ?? commandInfo.subcommand;
@@ -261,10 +258,9 @@ export function formatDiagnosticSummary(commandInfo, data) {
 function formatSessionText(data) {
     const sessions = getArrayField(data, "sessions");
     if (sessions) {
-        const visibleSessions = filterCallerOwnedSessionListItems(sessions);
-        if (visibleSessions.length === 0)
-            return sessions.length === 0 ? "No active sessions." : "No caller-owned active sessions.";
-        return visibleSessions
+        if (sessions.length === 0)
+            return "No active sessions.";
+        return sessions
             .map((item, index) => {
             if (!isRecord(item))
                 return `${index + 1}. ${stringifyModelFacing(item)}`;
@@ -932,10 +928,9 @@ function formatStateText(data, subcommand) {
     }
     const states = getArrayField(data, "states") ?? getArrayField(data, "files");
     if (states) {
-        const visibleStates = filterCallerOwnedStateListItems(states);
-        if (visibleStates.length === 0)
-            return "No caller-owned saved states.";
-        return visibleStates
+        if (states.length === 0)
+            return "No saved states.";
+        return states
             .map((item, index) => {
             if (!isRecord(item))
                 return `${index + 1}. ${redactModelFacingTextIfSensitive(stringifyModelFacing(item))}`;
@@ -979,9 +974,9 @@ export function redactPresentationData(commandInfo, data) {
     if (commandInfo.command === "storage")
         return redactStorageData(data);
     if (commandInfo.command === "session" && commandInfo.subcommand === "list")
-        return redactStructuredPresentationValue(filterManagedSessionListRows(data));
+        return redactStructuredPresentationValue(data);
     if (commandInfo.command === "state" && commandInfo.subcommand === "list")
-        return redactStructuredPresentationValue(filterManagedStateListRows(data));
+        return redactStructuredPresentationValue(data);
     if (commandInfo.command === "state" && commandInfo.subcommand === "show")
         return redactStatefulValues(data, new Set(["value"]));
     return redactStructuredPresentationValue(data);

package/dist/extensions/agent-browser/lib/results/presentation/navigation.js

@@ -1,4 +1,4 @@
-import { isNavigationObservableCommandName, isPageChangeSummaryCommand } from "../../command-taxonomy.js";
+import { isNavigationObservableCommandName, isOpenNavigationCommand, isPageChangeSummaryCommand } from "../../command-taxonomy.js";
 import { isRecord } from "../../parsing.js";
 import { detectConfirmationRequired } from "../confirmation.js";
 import { omitUpstreamLifecycle, redactModelFacingText, stringifyModelFacing } from "./common.js";
@@ -74,8 +74,8 @@ export function formatExtractionText(commandInfo, data) {
     const safeOrigin = origin ? redactModelFacingText(origin) : undefined;
     return safeOrigin && safeOrigin !== safeScalarResult ? `${safeScalarResult}\n\nOrigin: ${safeOrigin}` : safeScalarResult;
 }
-export function isNavigationObservableCommand(command) {
-    return isNavigationObservableCommandName(command);
+export function isNavigationObservableCommand(command, subcommand) {
+    return isNavigationObservableCommandName(command, subcommand);
 }
 function isNavigationSummary(value) {
     return isRecord(value) && (typeof value.title === "string" || typeof value.url === "string");
@@ -95,7 +95,7 @@ function getTopLevelNavigationSummary(data) {
 function getNormalizedNavigationSummary(summary) {
     const title = typeof summary?.title === "string" && summary.title.trim().length > 0 ? summary.title.trim() : undefined;
     const url = typeof summary?.url === "string" && summary.url.trim().length > 0 ? summary.url.trim() : undefined;
-    return title || url ? { title, url } : undefined;
+    return title || url ? { title, url, ...(typeof summary?.urlChanged === "boolean" ? { urlChanged: summary.urlChanged } : {}) } : undefined;
 }
 export function formatNavigationSummary(summary) {
     const normalized = getNormalizedNavigationSummary(summary);
@@ -109,20 +109,22 @@ export function buildPageChangeSummary(options) {
     const { artifacts, commandInfo, data, nextActions, savedFilePath } = options;
     const artifactCount = artifacts?.length ?? 0;
     const navigation = isRecord(data)
-        ? getNormalizedNavigationSummary(getNavigationSummary(data) ?? (isPageChangeSummaryCommand(commandInfo.command) ? getTopLevelNavigationSummary(data) : undefined))
+        ? getNormalizedNavigationSummary(getNavigationSummary(data) ?? (isPageChangeSummaryCommand(commandInfo.command, commandInfo.subcommand) ? getTopLevelNavigationSummary(data) : undefined))
         : undefined;
     const confirmationRequired = detectConfirmationRequired(data) !== undefined;
-    if (!navigation && !confirmationRequired && artifactCount === 0 && !savedFilePath && !isPageChangeSummaryCommand(commandInfo.command)) {
+    if (!navigation && !confirmationRequired && artifactCount === 0 && !savedFilePath && !isPageChangeSummaryCommand(commandInfo.command, commandInfo.subcommand)) {
         return undefined;
     }
+    const navigationObserved = navigation && (navigation.urlChanged === true || isOpenNavigationCommand(commandInfo.command) || ["back", "forward", "pushstate", "reload"].includes(commandInfo.command ?? ""));
     const changeType = savedFilePath || artifactCount > 0
         ? "artifact"
-        : navigation
+        : navigationObserved
             ? "navigation"
             : confirmationRequired
                 ? "confirmation"
                 : "mutation";
-    const parts = [commandInfo.command ?? "agent-browser", changeType];
+    const observed = changeType !== "mutation";
+    const parts = [commandInfo.command ?? "agent-browser", observed ? changeType : "action dispatched"];
     if (navigation?.title)
         parts.push(navigation.title);
     if (navigation?.url)
@@ -136,8 +138,9 @@ export function buildPageChangeSummary(options) {
         changeType,
         ...(commandInfo.command ? { command: commandInfo.command } : {}),
         ...(nextActions ? { nextActionIds: nextActions.map((action) => action.id) } : {}),
+        observed,
         ...(savedFilePath ? { savedFilePath } : {}),
-        summary: parts.join(" → "),
+        summary: `${parts.join(" → ")}${observed ? "" : " → application change unverified"}`,
         ...(navigation?.title ? { title: navigation.title } : {}),
         ...(navigation?.url ? { url: navigation.url } : {}),
     };