fix(vom): add record safe observations, unify definition and construction of safe DTO

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

@@ -3146,11 +3146,12 @@ describe("handleSnapshot", () => {
     expect(result.rootFrameId).toBe("root");
     expect(result).not.toHaveProperty("frameDocuments");
     expect(result.frames).toEqual([{ frameId: "root", target: { tabId: 4 } }]);
-    expect(result.nodes.find((node) => node.backendNodeId === 13)).toMatchObject({
+    expect(result.matchNodes.find((node) => node.backendNodeId === 13)).toMatchObject({
       frameId: "root",
       backendNodeId: 13,
       tag: "input",
       rect: { x: 400, y: 300, w: 200, h: 40 },
+      localRect: { x: 400, y: 300, w: 200, h: 40 },
     });
   });
 
@@ -3214,7 +3215,10 @@ describe("handleSnapshot", () => {
     expect(dumped).not.toContain(secrets.otp);
     expect(dumped).not.toContain(secrets.card);
     expect(dumped).not.toContain("formValue");
+    expect(dumped).not.toContain("formDefaultValue");
     expect(dumped).not.toContain("axNodes");
+    expect(dumped).not.toContain("domNodes");
+    expect(dumped).not.toContain("attrs");
     expect(result.text).toContain("•••");
   });
 
@@ -3231,9 +3235,10 @@ describe("handleSnapshot", () => {
       ownerBackendNodeId: 12,
       target: { tabId: 4, sessionId: "child-session" },
     });
-    expect(result.nodes.find((node) => node.backendNodeId === 22)).toMatchObject({
+    expect(result.matchNodes.find((node) => node.backendNodeId === 22)).toMatchObject({
       frameId: "child",
       tag: "button",
+      localRect: { x: 20, y: 30, w: 120, h: 40 },
     });
     expect(result.refs.find((ref) => ref.backendNodeId === 22)).toMatchObject({
       frameId: "child",

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

@@ -1,7 +1,9 @@
 export {
-  type CaptureVomFrame,
-  type CaptureVomNode,
   type CaptureVomObservationOptions,
-  type CaptureVomObservationResult,
   captureVomObservation,
 } from "./observation";
+export {
+  type CaptureVomFrame,
+  type CaptureVomMatchNode,
+  type CaptureVomObservationResult,
+} from "./vom/record-safe-observation";

apps/extension/src/tools/observation.ts

@@ -8,8 +8,6 @@ import {
   applyVomInteractionRecovery,
   type CondSurface,
   isVomReferenceNode,
-  type Rect,
-  type RenderedRef,
   renderVom,
   type VomNode,
   type VomOptions,
@@ -51,12 +49,15 @@ import {
 import { resolveSnapshotRef } from "./snapshot-ref";
 import {
   type CapturedNode,
-  type CapturedSurfaceProbe,
   type CapturedViewModel,
   captureViewModel,
   collectOverlayExcludedBackendIds,
 } from "./vom/capture";
 import { type CapturedFrameDocument, captureFrameData } from "./vom/frame-capture";
+import {
+  type CaptureVomObservationResult,
+  projectRecordSafeObservation,
+} from "./vom/record-safe-observation";
 
 // ---------------------------------------------------------------------------
 // Shared helpers (legacy aliases — observation.ts kept exporting these
@@ -1591,58 +1592,6 @@ export interface CaptureVomObservationOptions extends VomOptions {
   signal?: AbortSignal;
 }
 
-/** Frame identity needed to click an `@eN` that lives in an iframe. */
-export interface CaptureVomFrame {
-  frameId: string;
-  target: CdpTarget;
-  parentFrameId?: string;
-  ownerBackendNodeId?: number;
-}
-
-/** Geometry index for matching a recorded click to a captured node. */
-export interface CaptureVomNode {
-  frameId: string;
-  backendNodeId: number;
-  tag: string;
-  rect: Rect | null;
-}
-
-/**
- * Record-safe observation: rendered VOM text plus the matching index.
- * Raw AX/DOM trees are not returned — they carry form values.
- */
-export interface CaptureVomObservationResult {
-  text: string;
-  refs: RenderedRef[];
-  truncated: boolean;
-  rootFrameId: string;
-  frames: CaptureVomFrame[];
-  nodes: CaptureVomNode[];
-  surfaceProbes?: CapturedSurfaceProbe[];
-}
-
-function recordSafeFrames(documents: CapturedFrameDocument<CdpAxNode>[]): CaptureVomFrame[] {
-  return documents.map((document) => ({
-    frameId: document.frameId,
-    target: document.target,
-    ...(document.parentFrameId ? { parentFrameId: document.parentFrameId } : {}),
-    ...(document.ownerBackendNodeId !== undefined
-      ? { ownerBackendNodeId: document.ownerBackendNodeId }
-      : {}),
-  }));
-}
-
-function recordSafeNodes(documents: CapturedFrameDocument<CdpAxNode>[]): CaptureVomNode[] {
-  return documents.flatMap((document) =>
-    document.domNodes.map((node) => ({
-      frameId: document.frameId,
-      backendNodeId: node.backendNodeId,
-      tag: node.tag,
-      rect: node.rect,
-    })),
-  );
-}
-
 export async function captureVomObservation(
   cdp: CdpRunner,
   tabId: number,
@@ -1667,15 +1616,12 @@ export async function captureVomObservation(
     activeRegionPolicy: options.activeRegionPolicy,
   });
   throwIfAborted(options.signal, "observation");
-  return {
+  return projectRecordSafeObservation({
-    text: rendered.text,
-    refs: rendered.refs,
-    truncated: rendered.truncated,
     rootFrameId: captured.rootFrameId ?? documents[0]?.frameId ?? "root",
-    frames: recordSafeFrames(documents),
+    frameDocuments: documents,
-    nodes: recordSafeNodes(documents),
+    rendered,
     surfaceProbes: captured.surfaceProbes,
-  };
+  });
 }
 
 async function handleVomObservation(

apps/extension/src/tools/vom/record-safe-observation.ts

@@ -0,0 +1,77 @@
+import type { Rect, RenderedRef, VomResult } from "@browser-skill/vom";
+import type { CdpTarget } from "@/browser-driver/frame-graph";
+import type { CapturedSurfaceProbe } from "./capture";
+import type { CapturedFrameDocument, FrameAxNode } from "./frame-capture";
+
+/** Frame identity needed to resolve an `@eN` that lives in an iframe. */
+export interface CaptureVomFrame {
+  frameId: string;
+  target: CdpTarget;
+  parentFrameId?: string;
+  ownerBackendNodeId?: number;
+}
+
+/** Allowlisted geometry used to match a recorded action to a rendered ref. */
+export interface CaptureVomMatchNode {
+  frameId: string;
+  backendNodeId: number;
+  tag: string;
+  /** Top-level viewport-relative geometry. */
+  rect: Rect | null;
+  /** Frame-local viewport-relative geometry, when available. */
+  localRect?: Rect | null;
+}
+
+/**
+ * Raw AX/DOM trees are never returned. When `redactValues` is enabled by the
+ * caller, form values in `text` are masked as well.
+ */
+export interface CaptureVomObservationResult {
+  text: string;
+  refs: RenderedRef[];
+  truncated: boolean;
+  rootFrameId: string;
+  frames: CaptureVomFrame[];
+  matchNodes: CaptureVomMatchNode[];
+  surfaceProbes?: CapturedSurfaceProbe[];
+}
+
+function projectFrames(documents: CapturedFrameDocument<FrameAxNode>[]): CaptureVomFrame[] {
+  return documents.map((document) => ({
+    frameId: document.frameId,
+    target: document.target,
+    ...(document.parentFrameId ? { parentFrameId: document.parentFrameId } : {}),
+    ...(document.ownerBackendNodeId !== undefined
+      ? { ownerBackendNodeId: document.ownerBackendNodeId }
+      : {}),
+  }));
+}
+
+function projectMatchNodes(documents: CapturedFrameDocument<FrameAxNode>[]): CaptureVomMatchNode[] {
+  return documents.flatMap((document) =>
+    document.domNodes.map((node) => ({
+      frameId: document.frameId,
+      backendNodeId: node.backendNodeId,
+      tag: node.tag,
+      rect: node.rect,
+      ...(node.localRect !== undefined ? { localRect: node.localRect } : {}),
+    })),
+  );
+}
+
+export function projectRecordSafeObservation(input: {
+  rootFrameId: string;
+  frameDocuments: CapturedFrameDocument<FrameAxNode>[];
+  rendered: VomResult;
+  surfaceProbes?: CapturedSurfaceProbe[];
+}): CaptureVomObservationResult {
+  return {
+    text: input.rendered.text,
+    refs: input.rendered.refs,
+    truncated: input.rendered.truncated,
+    rootFrameId: input.rootFrameId,
+    frames: projectFrames(input.frameDocuments),
+    matchNodes: projectMatchNodes(input.frameDocuments),
+    surfaceProbes: input.surfaceProbes,
+  };
+}