diff --git a/src/media-understanding/runner.attachments.ts b/src/media-understanding/runner.attachments.ts
new file mode 100644
index 00000000000..b64804b8cfc
--- /dev/null
+++ b/src/media-understanding/runner.attachments.ts
@@ -0,0 +1,18 @@
+import type { MsgContext } from "../auto-reply/templating.js";
+import {
+  MediaAttachmentCache,
+  type MediaAttachmentCacheOptions,
+  normalizeAttachments,
+} from "./attachments.js";
+import type { MediaAttachment } from "./types.js";
+
+export function normalizeMediaAttachments(ctx: MsgContext): MediaAttachment[] {
+  return normalizeAttachments(ctx);
+}
+
+export function createMediaAttachmentCache(
+  attachments: MediaAttachment[],
+  options?: MediaAttachmentCacheOptions,
+): MediaAttachmentCache {
+  return new MediaAttachmentCache(attachments, options);
+}
diff --git a/src/media-understanding/runner.skip-tiny-audio.test.ts b/src/media-understanding/runner.skip-tiny-audio.test.ts
index a4021fb52a8..127b5546f22 100644
--- a/src/media-understanding/runner.skip-tiny-audio.test.ts
+++ b/src/media-understanding/runner.skip-tiny-audio.test.ts
@@ -5,12 +5,8 @@ import { describe, expect, it } from "vitest";
 import type { MsgContext } from "../auto-reply/templating.js";
 import type { OpenClawConfig } from "../config/config.js";
 import { MIN_AUDIO_FILE_BYTES } from "./defaults.js";
-import {
-  buildProviderRegistry,
-  createMediaAttachmentCache,
-  normalizeMediaAttachments,
-  runCapability,
-} from "./runner.js";
+import { createMediaAttachmentCache, normalizeMediaAttachments } from "./runner.attachments.js";
+import { buildProviderRegistry, runCapability } from "./runner.js";
 import type { AudioTranscriptionRequest } from "./types.js";
 
 async function withAudioFixture(params: {
diff --git a/src/media-understanding/runner.test-utils.ts b/src/media-understanding/runner.test-utils.ts
index 086418f049d..8b86b97b805 100644
--- a/src/media-understanding/runner.test-utils.ts
+++ b/src/media-understanding/runner.test-utils.ts
@@ -3,7 +3,7 @@ import os from "node:os";
 import path from "node:path";
 import { withEnvAsync } from "../test-utils/env.js";
 import { MIN_AUDIO_FILE_BYTES } from "./defaults.js";
-import { createMediaAttachmentCache, normalizeMediaAttachments } from "./runner.js";
+import { createMediaAttachmentCache, normalizeMediaAttachments } from "./runner.attachments.js";
 
 type MediaFixtureParams = {
   ctx: { MediaPath: string; MediaType: string };
diff --git a/src/media-understanding/runner.ts b/src/media-understanding/runner.ts
index fa9a7379e23..7eb82cacf43 100644
--- a/src/media-understanding/runner.ts
+++ b/src/media-understanding/runner.ts
@@ -25,12 +25,7 @@ import {
 } from "../media/inbound-path-policy.js";
 import { getDefaultMediaLocalRoots } from "../media/local-roots.js";
 import { runExec } from "../process/exec.js";
-import {
-  MediaAttachmentCache,
-  type MediaAttachmentCacheOptions,
-  normalizeAttachments,
-  selectAttachments,
-} from "./attachments.js";
+import { MediaAttachmentCache, selectAttachments } from "./attachments.js";
 import {
   AUTO_AUDIO_KEY_PROVIDERS,
   AUTO_IMAGE_KEY_PROVIDERS,
@@ -60,6 +55,7 @@ import type {
   MediaUnderstandingOutput,
   MediaUnderstandingProvider,
 } from "./types.js";
+export { createMediaAttachmentCache, normalizeMediaAttachments } from "./runner.attachments.js";
 
 export type ActiveMediaModel = {
   provider: string;
@@ -80,10 +76,6 @@ export function buildProviderRegistry(
   return buildMediaUnderstandingRegistry(overrides, cfg);
 }
 
-export function normalizeMediaAttachments(ctx: MsgContext): MediaAttachment[] {
-  return normalizeAttachments(ctx);
-}
-
 export function resolveMediaAttachmentLocalRoots(params: {
   cfg: OpenClawConfig;
   ctx: MsgContext;
@@ -97,13 +89,6 @@ export function resolveMediaAttachmentLocalRoots(params: {
   );
 }
 
-export function createMediaAttachmentCache(
-  attachments: MediaAttachment[],
-  options?: MediaAttachmentCacheOptions,
-): MediaAttachmentCache {
-  return new MediaAttachmentCache(attachments, options);
-}
-
 const binaryCache = new Map<string, Promise<string | null>>();
 const geminiProbeCache = new Map<string, Promise<boolean>>();
 
diff --git a/src/media-understanding/runner.vision-skip.test.ts b/src/media-understanding/runner.vision-skip.test.ts
index c671e8202ff..ba227510737 100644
--- a/src/media-understanding/runner.vision-skip.test.ts
+++ b/src/media-understanding/runner.vision-skip.test.ts
@@ -10,6 +10,7 @@ import { __testing as loaderTesting } from "../plugins/loader.js";
 import { loadPluginManifestRegistry } from "../plugins/manifest-registry.js";
 import { createEmptyPluginRegistry } from "../plugins/registry.js";
 import { setActivePluginRegistry } from "../plugins/runtime.js";
+import { createMediaAttachmentCache, normalizeMediaAttachments } from "./runner.attachments.js";
 
 const catalog = [
   {
@@ -33,8 +34,6 @@ vi.mock("../agents/model-catalog.js", async () => {
 });
 
 let buildProviderRegistry: typeof import("./runner.js").buildProviderRegistry;
-let createMediaAttachmentCache: typeof import("./runner.js").createMediaAttachmentCache;
-let normalizeMediaAttachments: typeof import("./runner.js").normalizeMediaAttachments;
 let resolveAutoImageModel: typeof import("./runner.js").resolveAutoImageModel;
 let runCapability: typeof import("./runner.js").runCapability;
 
@@ -73,13 +72,16 @@ function setCompatibleActiveMediaUnderstandingRegistry(
 
 describe("runCapability image skip", () => {
   beforeAll(async () => {
-    ({
-      buildProviderRegistry,
-      createMediaAttachmentCache,
-      normalizeMediaAttachments,
-      resolveAutoImageModel,
-      runCapability,
-    } = await import("./runner.js"));
+    vi.doMock("../agents/model-catalog.js", async () => {
+      const actual = await vi.importActual<typeof import("../agents/model-catalog.js")>(
+        "../agents/model-catalog.js",
+      );
+      return {
+        ...actual,
+        loadModelCatalog,
+      };
+    });
+    ({ buildProviderRegistry, resolveAutoImageModel, runCapability } = await import("./runner.js"));
   });
 
   beforeEach(() => {