Merge pull request #203 from Tencent/fix/capture-document-consistency

fix(observation): validate DOM identity across capture phases | 校验跨采集阶段的 DOM 身份一致性

apps/extension/src/browser-driver/__tests__/chromium-cdp.test.ts

@@ -29,6 +29,24 @@ function fakeApi() {
 }
 
 describe("ChromiumCdp", () => {
+  it("changes attachment identity only across real debugger attachments", async () => {
+    const { api, onEvent, onDetach } = fakeApi();
+    const cdp = new ChromiumCdp(api);
+    expect(cdp.getAttachmentId(4)).toBeUndefined();
+    await cdp.ensureAttached(4);
+    const first = cdp.getAttachmentId(4);
+    expect(first).toBeTruthy();
+    await cdp.ensureAttached(4);
+    onEvent.fire({ tabId: 4 }, "Target.attachedToTarget", { sessionId: "child" });
+    expect(cdp.getAttachmentId(4)).toBe(first);
+    onDetach.fire({ tabId: 4 }, "target_closed");
+    expect(cdp.getAttachmentId(4)).toBeUndefined();
+    await cdp.ensureAttached(4);
+    expect(cdp.getAttachmentId(4)).not.toBe(first);
+    await cdp.detach(4);
+    expect(cdp.getAttachmentId(4)).toBeUndefined();
+  });
+
   it("discovers multiple iframe targets and recursively routes nested OOPIF commands", async () => {
     const { api, onEvent } = fakeApi();
     (api.sendCommand as ReturnType<typeof vi.fn>).mockImplementation(

apps/extension/src/browser-driver/chromium-cdp.ts

@@ -164,6 +164,7 @@ async function settleBeforeDeadline(promises: Promise<void>[], deadline: number)
 export class ChromiumCdp {
   private readonly api: CdpDebuggerApi;
   private readonly attachedTabs = new Set<number>();
+  private readonly attachmentIds = new Map<number, string>();
   private readonly attachInFlight = new Map<number, Promise<void>>();
   private readonly detachInFlight = new Map<number, Promise<void>>();
   private readonly tabOwners = new Map<number, Set<string>>();
@@ -198,6 +199,11 @@ export class ChromiumCdp {
     this.bindFrameTargetHandler();
   }
 
+  /** Identity of this actual debugger attachment, independent of frame topology. */
+  getAttachmentId(tabId: number): string | undefined {
+    return this.attachmentIds.get(tabId);
+  }
+
   /** Attach to `tabId` if we haven't already in this driver. */
   async ensureAttached(tabId: number): Promise<void> {
     // Returning a tab clears the cache before Chrome finishes detaching.
@@ -217,6 +223,7 @@ export class ChromiumCdp {
         await this.enableConsoleDomains(tabId);
         await this.enableNetworkDomainBestEffort(tabId);
         this.attachedTabs.add(tabId);
+        this.attachmentIds.set(tabId, crypto.randomUUID());
         await this.enableFrameDiscovery({ tabId }).catch((err) => {
           console.debug("[bsk cdp] frame discovery unavailable", { tabId, err });
         });
@@ -444,6 +451,7 @@ export class ChromiumCdp {
     this.attachInFlight.delete(tabId);
     if (!this.attachedTabs.has(tabId)) return;
     this.attachedTabs.delete(tabId);
+    this.attachmentIds.delete(tabId);
     this.clearDialogState(tabId);
     this.clearConsoleState(tabId);
     this.clearNetworkState(tabId);
@@ -502,6 +510,7 @@ export class ChromiumCdp {
     this.attachInFlight.clear();
     this.tabOwners.clear();
     this.attachedTabs.clear();
+    this.attachmentIds.clear();
     this.dialogBuffers.clear();
     this.dialogSequences.clear();
     this.consoleBuffers.clear();
@@ -848,6 +857,7 @@ export class ChromiumCdp {
     const listener = (source: chrome.debugger.Debuggee, _reason: string) => {
       if (typeof source.tabId === "number") {
         this.attachedTabs.delete(source.tabId);
+        this.attachmentIds.delete(source.tabId);
         this.attachInFlight.delete(source.tabId);
         this.tabOwners.delete(source.tabId);
         this.clearDialogState(source.tabId);

apps/extension/src/tools/__tests__/observation.test.ts

@@ -3695,7 +3695,8 @@ describe("handleSnapshot", () => {
     const result = await captureVomObservation(makeFrameAwareDeps().cdp, 4, "https://example.com", {
       maxTokens,
     });
-    expect(result.text.match(/@warning/g)).toHaveLength(1);
+    expect(result.text.match(/@warning/g)).toHaveLength(2);
+    expect(result.text).toContain("document identity unverified");
     expect(result.text).toContain("geometry incomplete");
     expect(result.text.startsWith("@vom 1\n")).toBe(true);
     if (maxTokens === undefined) {
@@ -3722,7 +3723,8 @@ describe("handleSnapshot", () => {
     for (const maxTokens of [undefined, 100, 0]) {
       const result = await captureVomObservation(cdp, 4, "https://example.com", { maxTokens });
       expect(result.text).toContain(`@warning observation incomplete: some ${stage} data`);
-      expect(result.text.match(/@warning/g)).toHaveLength(1);
+      expect(result.text.match(/@warning/g)).toHaveLength(2);
+      expect(result.text).toContain("document identity unverified");
       expect(result.text.startsWith("@vom 1\n")).toBe(true);
       if (maxTokens === undefined) {
         expect(result.frames.map((frame) => frame.frameId)).toEqual(["main", "child"]);
@@ -3770,7 +3772,8 @@ describe("handleSnapshot", () => {
       4,
       "https://example.com",
     );
-    expect(result.text).not.toContain("@warning");
+    expect(result.text).not.toContain("iframe geometry incomplete");
+    expect(result.text).toContain("document identity unverified");
     const node = result.matchNodes.find((item) => item.backendNodeId === 22);
     if (ownerGeometry === "clipped") expect(node?.rect).toBeNull();
     else expect(node?.rect).toEqual({ x: 120, y: 130, w: 120, h: 40 });

apps/extension/src/tools/__tests__/snapshot-coordinates.browser.test.ts

@@ -130,6 +130,7 @@ describe.skipIf(!process.env.BSK_GEOMETRY_CHROME)("real DOMSnapshot coordinate c
               return send(method, params, sessionFor(target));
             },
             getFrameGraph: async () => graph,
+            getAttachmentId: () => rootSession,
           };
           const oracles = new Map<string, Oracle>();
           for (const frame of frames) {

apps/extension/src/tools/observation.ts

@@ -938,6 +938,18 @@ export async function captureVomObservation(
     notices.push(
       `@warning observation incomplete: some ${incompleteStages.join(", ")} data is unavailable or omitted.`,
     );
+  if (
+    facts.issues.some(
+      (issue) => issue.stage === "identity" && issue.reason !== "identity-unverified",
+    )
+  )
+    notices.push(
+      "@warning observation incomplete: some documents were omitted because their identity changed or could not be revalidated.",
+    );
+  if (facts.issues.some((issue) => issue.reason === "identity-unverified"))
+    notices.push(
+      "@warning document identity unverified: some retained documents could not be checked for changes during capture.",
+    );
   const captureNotice = notices.join("\n");
   const rendered = renderVom(decoratedScene, {
     maxDepth: options.maxDepth,

apps/extension/src/tools/shared.ts

@@ -55,6 +55,7 @@ export interface CdpRunner {
   sendToTarget?<T = unknown>(target: CdpTarget, method: string, params?: object): Promise<T>;
   detach?(tabId: number): Promise<void>;
   getFrameGraph?(tabId: number): Promise<CdpFrameGraph>;
+  getAttachmentId?(tabId: number): string | undefined;
   ensureAttachedToUrl?(tabId: number, expectedUrl: string | undefined): Promise<void>;
   trackSessionTab?(sessionId: string, tabId: number): void;
   releaseSessionTab?(sessionId: string, tabId: number): Promise<void>;

apps/extension/src/tools/vom/__tests__/capture-coordinator.test.ts

@@ -2,13 +2,22 @@ import { describe, expect, it, vi } from "vitest";
 import type { CdpFrame, CdpTarget } from "@/browser-driver/frame-graph";
 import { cdpTargetKey } from "@/browser-driver/frame-graph";
 import { OVERLAY_HOST_MARKER_ATTR } from "@/lib/overlay-bridge";
+import { captureVomObservation } from "../../observation";
 import type { CdpRunner } from "../../shared";
 import { captureObservationFacts, semanticCapture } from "../capture-coordinator";
 import { buildSemanticGraph } from "../semantic-graph/build";
 import { REQUESTED_STYLES, type SnapshotReply } from "../snapshot";
 
 function fixture(
-  options: { frames?: CdpFrame[]; fail?: string; omitDocument?: string; overlay?: string } = {},
+  options: {
+    frames?: CdpFrame[];
+    after?: Record<string, { element?: number; missing?: boolean; unreadable?: boolean }>;
+    fail?: string;
+    missingIdentity?: boolean;
+    omitDocument?: string;
+    attachmentChanged?: boolean;
+    overlay?: string;
+  } = {},
 ) {
   const frames = options.frames ?? [
     { frameId: "main", target: { tabId: 4 } },
@@ -34,7 +43,10 @@ function fixture(
   const elements = new Map(
     frames.map((frame, i) => [frame.frameId, frame.target.sessionId ? 1 : i * 10 + 1]),
   );
+  const contexts = new Map(frames.map((frame, i) => [i + 1, frame.frameId]));
   const logs: Array<{ target: CdpTarget; method: string; params: Record<string, unknown> }> = [];
+  let snapshots = 0;
+  const targetCount = new Set(frames.map((frame) => cdpTargetKey(frame.target))).size;
   const send = vi.fn(
     async <T>(target: CdpTarget, method: string, params: object = {}): Promise<T> => {
       logs.push({ target, method, params: params as Record<string, unknown> });
@@ -42,6 +54,27 @@ function fixture(
       if (options.fail === `${target.sessionId ?? "main"}:${method}`)
         throw new Error("fixture failure");
       let result: unknown = {};
+      if (method === "Page.createIsolatedWorld")
+        result = {
+          executionContextId: frames.findIndex((frame) => frame.frameId === args.frameId) + 1,
+        };
+      if (method === "Runtime.evaluate" && args.contextId) {
+        const id = contexts.get(Number(args.contextId))!;
+        const change = options.after?.[id];
+        result =
+          options.missingIdentity || change?.unreadable
+            ? {}
+            : {
+                result: {
+                  deepSerializedValue: change?.missing
+                    ? { type: "null" }
+                    : {
+                        type: "node",
+                        value: { backendNodeId: change?.element ?? elements.get(id) },
+                      },
+                },
+              };
+      }
       if (method === "Page.getLayoutMetrics")
         result = {
           visualViewport: { clientWidth: 1000 },
@@ -49,6 +82,7 @@ function fixture(
           cssLayoutViewport: { clientWidth: 1000, clientHeight: 800 },
         };
       if (method === "DOMSnapshot.captureSnapshot") {
+        snapshots++;
         expect((params as { computedStyles: unknown }).computedStyles).toEqual(REQUESTED_STYLES);
         result = {
           strings: [
@@ -137,6 +171,12 @@ function fixture(
     },
   );
   const cdp: CdpRunner = {
+    getAttachmentId: () =>
+      options.missingIdentity
+        ? undefined
+        : options.attachmentChanged && snapshots === targetCount
+          ? "new-attachment"
+          : "attachment",
     getFrameGraph: async () => ({ rootFrameId: frames[0].frameId, frames }),
     send: (tabId, method, params) =>
       (send as NonNullable<CdpRunner["sendToTarget"]>)({ tabId }, method, params),
@@ -181,6 +221,9 @@ describe("captureObservationFacts", () => {
     expect(logs.filter((call) => call.method === "DOMSnapshot.captureSnapshot")).toHaveLength(2);
     expect(logs.filter((call) => call.method === "Accessibility.enable")).toHaveLength(2);
     expect(logs.filter((call) => call.method === "Accessibility.getFullAXTree")).toHaveLength(4);
+    expect(facts.documents.every((doc) => doc.identity)).toBe(true);
+    expect(logs.filter((call) => call.method === "DOM.describeNode")).toHaveLength(0);
+    expect(logs.filter((call) => call.method === "Page.createIsolatedWorld")).toHaveLength(4);
     const main = facts.documents.find((doc) => doc.frame.frameId === "main")!;
     const remote = facts.documents.find((doc) => doc.frame.frameId === "remote")!;
     expect(main.index.nodes.get(2)).not.toBe(remote.index.nodes.get(2));
@@ -191,6 +234,219 @@ describe("captureObservationFacts", () => {
     expect(facts.finishedAt).toBeGreaterThanOrEqual(facts.startedAt);
   });
 
+  it.each([
+    { element: 99 },
+    { missing: true },
+  ])("isolates changed child identity %j and dependent descendants", async (change) => {
+    const { cdp } = fixture({ after: { same: change } });
+    const facts = await captureObservationFacts(cdp, 4);
+    expect(facts.documents.map((doc) => doc.frame.frameId)).toEqual(["main", "remote"]);
+    expect(facts.issues).toContainEqual(
+      expect.objectContaining({ frameId: "same", reason: "document-changed" }),
+    );
+  });
+
+  it.each([
+    { after: { main: { element: 99 } } },
+    { attachmentChanged: true },
+  ])("rejects a root replacement or reattachment %j", async (options) => {
+    await expect(captureObservationFacts(fixture(options).cdp, 4)).rejects.toThrow(
+      "document identity",
+    );
+  });
+
+  it.each([
+    "main",
+    "same",
+  ])("does not publish a previously identified %s document after identity reads fail", async (frameId) => {
+    const { cdp } = fixture({ after: { [frameId]: { unreadable: true } } });
+    if (frameId === "main")
+      await expect(captureObservationFacts(cdp, 4)).rejects.toThrow("document identity");
+    else {
+      const facts = await captureObservationFacts(cdp, 4);
+      expect(facts.documents.map((doc) => doc.frame.frameId)).toEqual(["main", "remote"]);
+      expect(facts.issues).toContainEqual(
+        expect.objectContaining({ frameId, reason: "identity-unavailable" }),
+      );
+    }
+  });
+
+  it("validates newly discovered snapshot documents without a pre-snapshot identity read", async () => {
+    const { cdp, logs } = fixture();
+    const graph = await cdp.getFrameGraph!(4);
+    cdp.getFrameGraph = async () => ({ ...graph, frames: graph.frames.slice(0, 1) });
+    const facts = await captureObservationFacts(cdp, 4);
+    expect(facts.documents.map((doc) => doc.frame.frameId)).toEqual(["main", "same", "nested"]);
+    expect(facts.documents.every((doc) => doc.identity)).toBe(true);
+    expect(logs.filter((call) => call.method === "Page.createIsolatedWorld")).toHaveLength(3);
+    expect(logs.findIndex((call) => call.method === "Page.createIsolatedWorld")).toBeGreaterThan(
+      logs.findIndex((call) => call.method === "Accessibility.getFullAXTree"),
+    );
+  });
+
+  it("rejects a snapshot root that no longer matches the current frame root", async () => {
+    const { cdp } = fixture();
+    const original = cdp.sendToTarget!;
+    cdp.sendToTarget = async (target, method, params) => {
+      const result = await original(target, method, params);
+      if (method === "DOMSnapshot.captureSnapshot" && !target.sessionId) {
+        const snapshot = result as {
+          documents: { frameId: string; nodes: { backendNodeId: number[] } }[];
+        };
+        snapshot.documents[0].nodes.backendNodeId[2] = 999;
+      }
+      return result as never;
+    };
+    cdp.send = (tabId, method, params) => cdp.sendToTarget!({ tabId }, method, params);
+    await expect(captureObservationFacts(cdp, 4)).rejects.toThrow("document identity");
+  });
+
+  it("starts after-identity reads only after every target finishes and propagates cancellation during identity cleanup", async () => {
+    const { cdp, logs } = fixture();
+    const controller = new AbortController();
+    const original = cdp.sendToTarget!;
+    let identityCleanup = 0;
+    cdp.sendToTarget = async (target, method, params) => {
+      if (method === "Page.createIsolatedWorld") {
+        expect(logs.filter((call) => call.method === "Accessibility.getFullAXTree")).toHaveLength(
+          4,
+        );
+      }
+      const result = await original(target, method, params);
+      if (
+        method === "Runtime.releaseObjectGroup" &&
+        String((params as { objectGroup?: string })?.objectGroup).startsWith(
+          "bsk-document-identity-",
+        ) &&
+        target.sessionId === "remote"
+      ) {
+        if (++identityCleanup === 1) controller.abort();
+      }
+      return result as never;
+    };
+    cdp.send = (tabId, method, params) => cdp.sendToTarget!({ tabId }, method, params);
+    await expect(captureObservationFacts(cdp, 4, controller.signal)).rejects.toMatchObject({
+      name: "AbortError",
+    });
+    expect(identityCleanup).toBe(1);
+  });
+
+  it.each([
+    1, 10, 100,
+  ])("bounds identity reads for %i same-target documents after snapshot and AX collection", async (count) => {
+    const frames: CdpFrame[] = Array.from({ length: count }, (_, i) => ({
+      frameId: `frame-${i}`,
+      target: { tabId: 4 },
+      ...(i ? { parentFrameId: "frame-0", ownerBackendNodeId: 90000 + i } : {}),
+    }));
+    const { cdp, logs } = fixture({ frames });
+    const original = cdp.sendToTarget!;
+    let active = 0;
+    let peak = 0;
+    let released = 0;
+    cdp.sendToTarget = async (target, method, params) => {
+      if (method === "Page.createIsolatedWorld") {
+        peak = Math.max(peak, ++active);
+        await new Promise((resolve) => setTimeout(resolve, 1));
+      }
+      if (method === "DOMSnapshot.captureSnapshot") {
+        expect(active).toBe(0);
+        expect(released).toBe(0);
+      }
+      if (method === "Page.createIsolatedWorld") {
+        expect(logs.filter((call) => call.method === "Accessibility.getFullAXTree")).toHaveLength(
+          count,
+        );
+      }
+      const result = await original(target, method, params);
+      if (
+        method === "Runtime.releaseObjectGroup" &&
+        String((params as { objectGroup?: string })?.objectGroup).startsWith(
+          "bsk-document-identity-",
+        )
+      ) {
+        active--;
+        released++;
+      }
+      return result as never;
+    };
+    cdp.send = (tabId, method, params) => cdp.sendToTarget!({ tabId }, method, params);
+    const facts = await captureObservationFacts(cdp, 4);
+    expect(facts.documents.every((doc) => doc.identity)).toBe(true);
+    expect(peak).toBe(Math.min(count, 4));
+    expect(active).toBe(0);
+    expect(released).toBe(count);
+    for (const method of [
+      "Page.createIsolatedWorld",
+      "Runtime.evaluate",
+      "Runtime.releaseObjectGroup",
+    ])
+      expect(logs.filter((call) => call.method === method)).toHaveLength(count);
+    expect(logs.filter((call) => call.method === "DOMSnapshot.captureSnapshot")).toHaveLength(1);
+    expect(logs.filter((call) => call.method === "Page.getFrameTree")).toHaveLength(0);
+  });
+
+  it("joins identity cleanup on cancellation without starting queued frames", async () => {
+    const frames: CdpFrame[] = Array.from({ length: 5 }, (_, i) => ({
+      frameId: `frame-${i}`,
+      target: { tabId: 4 },
+      ...(i ? { parentFrameId: "frame-0", ownerBackendNodeId: 90000 + i } : {}),
+    }));
+    const { cdp, logs } = fixture({ frames });
+    const original = cdp.sendToTarget!;
+    const controller = new AbortController();
+    let release!: () => void;
+    const gate = new Promise<void>((resolve) => {
+      release = resolve;
+    });
+    let ready!: () => void;
+    const blocked = new Promise<void>((resolve) => {
+      ready = resolve;
+    });
+    let held = 0;
+    cdp.sendToTarget = async (target, method, params) => {
+      const result = await original(target, method, params);
+      if (
+        method === "Runtime.releaseObjectGroup" &&
+        String((params as { objectGroup?: string })?.objectGroup).startsWith(
+          "bsk-document-identity-",
+        )
+      ) {
+        if (++held === 4) ready();
+        await gate;
+      }
+      return result as never;
+    };
+    cdp.send = (tabId, method, params) => cdp.sendToTarget!({ tabId }, method, params);
+    let settled = false;
+    const pending = captureObservationFacts(cdp, 4, controller.signal).finally(() => {
+      settled = true;
+    });
+    const rejected = expect(pending).rejects.toMatchObject({ name: "AbortError" });
+    await blocked;
+    controller.abort();
+    await Promise.resolve();
+    expect(settled).toBe(false);
+    release();
+    await rejected;
+    expect(held).toBe(4);
+    expect(logs.filter((call) => call.method === "Page.createIsolatedWorld")).toHaveLength(4);
+  });
+
+  it.each([
+    [{ after: { same: { element: 99 } } }, "omitted", false],
+    [{ after: { same: { unreadable: true } } }, "omitted", false],
+    [{ missingIdentity: true }, "document identity unverified", true],
+    [{}, "", true],
+  ] as const)("publishes identity integrity without silently losing content %j", async (options, notice, retained) => {
+    const { cdp } = fixture(options);
+    const result = await captureVomObservation(cdp, 4, "https://example.com");
+    expect(result.frames.some((frame) => frame.frameId === "same")).toBe(retained);
+    if (notice) expect(result.text).toContain(notice);
+    else expect(result.text).not.toContain("identity");
+    if (!retained) expect(result.text).not.toContain("document identity unverified");
+  });
+
   it("does not retry failed or missing documents and retains valid AX-only semantics", async () => {
     const { cdp, logs } = fixture({
       fail: "remote:DOMSnapshot.captureSnapshot",
@@ -209,6 +465,12 @@ describe("captureObservationFacts", () => {
     );
   });
 
+  it("keeps missing identity explicit without manufacturing a verified document", async () => {
+    const facts = await captureObservationFacts(fixture({ missingIdentity: true }).cdp, 4);
+    expect(facts.documents.every((doc) => !doc.identity)).toBe(true);
+    expect(facts.issues.filter((issue) => issue.reason === "identity-unverified")).toHaveLength(4);
+  });
+
   it("excludes overlay AX in its own document without excluding equal IDs elsewhere", async () => {
     const facts = await captureObservationFacts(fixture({ overlay: "remote" }).cdp, 4);
     const { documents } = semanticCapture(facts);
@@ -906,7 +1168,7 @@ describe("sibling frame measurement scheduling", () => {
       return original(tabId, method, params);
     };
     const captured = await captureObservationFacts(fixture.cdp, 4);
-    expect(captured.issues).toEqual([]);
+    expect(captured.issues.filter((issue) => issue.stage === "geometry")).toEqual([]);
     expect(captured.documents.map((doc) => doc.frame.frameId)).toEqual(
       Array.from({ length: 8 }, (_, i) => `frame-${i}`),
     );

apps/extension/src/tools/vom/capture-coordinator.ts

@@ -11,7 +11,8 @@ import {
   isAbortError as isCaptureAbort,
   throwIfAborted as throwCaptureAborted,
 } from "./capture-abort";
-import type { CapturedSceneInput } from "./facts";
+import { verifyDocumentIdentity } from "./document-identity";
+import type { CapturedSceneInput, DocumentIdentity } from "./facts";
 import { buildDocumentIndex, type CaptureIssue, type ObservationFacts } from "./facts";
 import { enrichFormControlStates } from "./form-capture";
 import {
@@ -26,6 +27,7 @@ import { describeSnapshotFrames, REQUESTED_STYLES, type SnapshotReply } from "./
 interface TargetBatch<T extends FrameOwnedAxNode> {
   target: CdpTarget;
   frames: CdpFrame[];
+  snapshotAttachmentId?: string;
   documents: NormalizedFrameDocument[];
   ax: FrameAxBatch<T>[];
   fallbackExcluded: Set<number>;
@@ -121,6 +123,7 @@ export async function captureObservationFacts<T extends FrameOwnedAxNode>(
         throwCaptureAborted(signal);
         await scoped.send(tabId, "DOMSnapshot.enable", {});
         throwCaptureAborted(signal);
+        batch.snapshotAttachmentId = cdp.getAttachmentId?.(tabId);
         const snapshot = await scoped.send<SnapshotReply>(tabId, "DOMSnapshot.captureSnapshot", {
           computedStyles: REQUESTED_STYLES,
           includePaintOrder: true,
@@ -246,6 +249,53 @@ export async function captureObservationFacts<T extends FrameOwnedAxNode>(
   const rootBatch = batches.find((batch) => !batch.target.sessionId);
   const rootFrameId =
     rootBatch?.rootFrameId ?? graph?.rootFrameId ?? rootBatch?.frames[0]?.frameId ?? "root";
+  const ownershipInvalid = new Set(invalid);
+  const identities = new Map<string, DocumentIdentity>();
+  const selected = new Map(currentFrames.map((frame) => [frame.frameId, frame]));
+  const checks = batches.flatMap((batch) => {
+    const docs = new Map(batch.documents.map((doc) => [doc.frame.frameId, doc]));
+    return batch.frames
+      .filter(
+        (frame) =>
+          !invalid.has(frame.frameId) &&
+          cdpTargetKey(selected.get(frame.frameId)!.target) === cdpTargetKey(batch.target),
+      )
+      .map((frame) => ({ batch, frame, doc: docs.get(frame.frameId) }));
+  });
+  await collectTasks(
+    checks,
+    async ({ batch, frame, doc }) => {
+      throwCaptureAborted(signal);
+      if (!batch.snapshotAttachmentId || doc?.documentElementBackendNodeId === undefined) {
+        issues.push({
+          target: frame.target,
+          frameId: frame.frameId,
+          stage: "identity",
+          reason: "identity-unverified",
+        });
+        return;
+      }
+      const identity: DocumentIdentity = {
+        attachmentId: batch.snapshotAttachmentId,
+        target: frame.target,
+        frameId: frame.frameId,
+        documentElementBackendNodeId: doc.documentElementBackendNodeId,
+      };
+      const verified = await verifyDocumentIdentity(cdp, identity, signal);
+      const status = cdp.getAttachmentId?.(tabId) === identity.attachmentId ? verified : "changed";
+      if (status === "current") identities.set(frame.frameId, identity);
+      else {
+        invalid.add(frame.frameId);
+        issues.push({
+          target: frame.target,
+          frameId: frame.frameId,
+          stage: "identity",
+          reason: status === "changed" ? "document-changed" : "identity-unavailable",
+        });
+      }
+    },
+    signal,
+  );
   const children = new Map<string, string[]>();
   for (const frame of currentFrames) {
     if (!frame.parentFrameId) continue;
@@ -263,7 +313,7 @@ export async function captureObservationFacts<T extends FrameOwnedAxNode>(
     }
   }
   for (const frame of currentFrames) {
-    if (invalid.has(frame.frameId))
+    if (ownershipInvalid.has(frame.frameId))
       issues.push({
         target: frame.target,
         frameId: frame.frameId,
@@ -271,7 +321,12 @@ export async function captureObservationFacts<T extends FrameOwnedAxNode>(
         reason: "frame-ownership-unresolved",
       });
   }
-  if (invalid.has(rootFrameId)) throw new Error("observation root document ownership is ambiguous");
+  if (invalid.has(rootFrameId))
+    throw new Error(
+      ownershipInvalid.has(rootFrameId)
+        ? "observation root document ownership is ambiguous"
+        : "observation document identity changed or could not be verified; observe again",
+    );
   const frames = currentFrames.filter((frame) => !invalid.has(frame.frameId));
   const frameById = new Map(frames.map((frame) => [frame.frameId, frame]));
   const belongs = (frame: CdpFrame) => {
@@ -319,6 +374,7 @@ export async function captureObservationFacts<T extends FrameOwnedAxNode>(
     } = document;
     facts.push({
       frame,
+      identity: identities.get(document.frameId),
       index: fallback?.size ? { ...index, excludedBackendNodeIds: fallback } : index,
       domNodes,
       axNodes,

apps/extension/src/tools/vom/document-identity.ts

@@ -0,0 +1,53 @@
+import type { CdpRunner } from "../shared";
+import { sendToCdpTarget } from "../shared";
+import { isAbortError, throwIfAborted } from "./capture-abort";
+import type { DocumentIdentity } from "./facts";
+
+/** Compare the snapshot root with the current root in that exact frame.
+ * Deep serialization supplies the backend ID without a separate describeNode. */
+export async function verifyDocumentIdentity(
+  cdp: CdpRunner,
+  identity: DocumentIdentity,
+  signal?: AbortSignal,
+): Promise<"current" | "changed" | "unavailable"> {
+  const attached = () => cdp.getAttachmentId?.(identity.target.tabId) === identity.attachmentId;
+  throwIfAborted(signal);
+  if (!attached()) return "changed";
+  const objectGroup = `bsk-document-identity-${crypto.randomUUID()}`;
+  const send = <T>(method: string, params: object) => {
+    throwIfAborted(signal);
+    return sendToCdpTarget<T>(cdp, identity.target, method, params);
+  };
+  try {
+    const world = await send<{ executionContextId: number }>("Page.createIsolatedWorld", {
+      frameId: identity.frameId,
+      worldName: "bsk-document-identity",
+    });
+    const reply = await send<{
+      result?: { deepSerializedValue?: { type: string; value?: { backendNodeId?: number } } };
+    }>("Runtime.evaluate", {
+      expression: "document.documentElement",
+      contextId: world.executionContextId,
+      objectGroup,
+      serializationOptions: {
+        serialization: "deep",
+        additionalParameters: { maxNodeDepth: 0, includeShadowTree: "none" },
+      },
+    });
+    if (!attached()) return "changed";
+    const root = reply.result?.deepSerializedValue;
+    if (root?.type === "null") return "changed";
+    const id = root?.type === "node" ? root.value?.backendNodeId : undefined;
+    if (id === undefined) return "unavailable";
+    return id === identity.documentElementBackendNodeId ? "current" : "changed";
+  } catch (error) {
+    throwIfAborted(signal);
+    if (isAbortError(error)) throw error;
+    return attached() ? "unavailable" : "changed";
+  } finally {
+    await sendToCdpTarget(cdp, identity.target, "Runtime.releaseObjectGroup", {
+      objectGroup,
+    }).catch(() => {});
+    throwIfAborted(signal);
+  }
+}

apps/extension/src/tools/vom/facts.ts

@@ -97,6 +97,13 @@ export async function buildDocumentIndex<T extends DecodedNode>(
   return { nodes, excludedBackendNodeIds };
 }
 
+export interface DocumentIdentity {
+  attachmentId: string;
+  target: CdpTarget;
+  frameId: string;
+  documentElementBackendNodeId: number;
+}
+
 /** Narrow input shared by the existing semantic scene/hover consumers. */
 export interface CapturedSceneInput {
   nodes: CapturedNode[];
@@ -110,12 +117,19 @@ export interface CaptureIssue {
   projectionIssue?: FrameProjectionIssue;
   target: CdpTarget;
   frameId?: string;
-  stage: "dom" | "ax" | "ownership" | "geometry" | "forms";
-  reason: "capture-unavailable" | "frame-ownership-unresolved" | "geometry-unavailable";
+  stage: "dom" | "ax" | "identity" | "ownership" | "geometry" | "forms";
+  reason:
+    | "document-changed"
+    | "identity-unavailable"
+    | "identity-unverified"
+    | "capture-unavailable"
+    | "frame-ownership-unresolved"
+    | "geometry-unavailable";
 }
 
 export interface DocumentFacts<T extends FrameOwnedAxNode> {
   readonly frame: CdpFrame;
+  readonly identity?: DocumentIdentity;
   readonly index: DocumentIndex;
   readonly domNodes: CapturedNode[];
   readonly axNodes: T[];

apps/extension/src/tools/vom/normalize.ts

@@ -38,6 +38,7 @@ export interface FrameContext {
 export interface NormalizedDocument {
   nodes: NodeFacts[];
   index: DocumentIndex;
+  documentElementBackendNodeId?: number;
 }
 
 /** Interpret one document using a supplied projection. No live reads or frame scheduling. */
@@ -98,6 +99,12 @@ export async function normalizeDocument(
       (node) => !node.tag.startsWith("#") && !index.excludedBackendNodeIds.has(node.backendNodeId),
     ),
     index,
+    documentElementBackendNodeId: nodes.find(
+      (node) =>
+        node.nodeType === 1 &&
+        node.parentBackendNodeId !== null &&
+        index.nodes.get(node.parentBackendNodeId)?.nodeType === 9,
+    )?.backendNodeId,
   };
 }