diff --git a/core-web/libs/edit-content-bridge/src/lib/bridges/angular-form-bridge.spec.ts b/core-web/libs/edit-content-bridge/src/lib/bridges/angular-form-bridge.spec.ts
index 279276694eca..46586ca0bdf1 100644
--- a/core-web/libs/edit-content-bridge/src/lib/bridges/angular-form-bridge.spec.ts
+++ b/core-web/libs/edit-content-bridge/src/lib/bridges/angular-form-bridge.spec.ts
@@ -8,6 +8,8 @@ import { DotAssetPickerComponent } from '@dotcms/ui';
 
 import { AngularFormBridge } from './angular-form-bridge';
 
+import { DotBrowserOptions } from '../interfaces/asset-browser.interface';
+
 /** The site the picker browses; the bridge is handed a way to resolve it. */
 const SITE: DotSite = {
     identifier: 'site-1',
@@ -942,7 +944,6 @@ describe('AngularFormBridge', () => {
                 bridge.openBrowserModal();
 
                 expect(openedConfig().allowedBaseTypes).toEqual(['DOTASSET', 'FILEASSET']);
-                expect(openedConfig().browse?.showFolders).toBeFalsy();
                 expect(openedConfig().browse?.showLinks).toBeFalsy();
             });
         });
@@ -954,12 +955,66 @@ describe('AngularFormBridge', () => {
                 expect(openedConfig().allowedBaseTypes).toEqual(['FILEASSET', 'HTMLPAGE']);
             });
 
-            it('should map folder and link kinds to browse options', () => {
-                bridge.openBrowserModal({ kinds: ['page', 'folder', 'link'] });
+            it('should map the link kind to a browse option', () => {
+                bridge.openBrowserModal({ kinds: ['page', 'link'] });
 
-                expect(openedConfig().browse).toEqual(
-                    expect.objectContaining({ showFolders: true, showLinks: true })
-                );
+                expect(openedConfig().browse).toEqual(expect.objectContaining({ showLinks: true }));
+            });
+
+            it('should not carry a folder browse option for a caller that asks for folders', () => {
+                // #37366: `'folder'` left the contract, but a VTL template is a string literal —
+                // TypeScript polices nothing here, so the runtime has to. The kind is dropped, and
+                // the picker is never handed an option that would list folders.
+                bridge.openBrowserModal({
+                    kinds: ['page', 'folder', 'link']
+                } as unknown as DotBrowserOptions);
+
+                expect(openedConfig().browse).not.toHaveProperty('showFolders');
+                expect(openedConfig().browse).toEqual(expect.objectContaining({ showLinks: true }));
+            });
+
+            it('should warn about an unsupported kind rather than ignore it silently', () => {
+                // AC-008: a template author must not be able to ask for a kind the picker refuses
+                // and get no signal. Same treatment the `link` + `mimeTypes` conflict already gets.
+                const warn = jest.spyOn(console, 'warn').mockImplementation();
+
+                bridge.openBrowserModal({
+                    kinds: ['file', 'page', 'folder']
+                } as unknown as DotBrowserOptions);
+
+                expect(warn).toHaveBeenCalledTimes(1);
+                expect(warn.mock.calls[0][0]).toContain('folder');
+                expect(openedConfig().allowedBaseTypes).toEqual(['FILEASSET', 'HTMLPAGE']);
+
+                warn.mockRestore();
+            });
+
+            it('should fall back to asset-only browsing when folder is the only kind asked for', () => {
+                // Degenerate case: no requested kind maps to a base type, so `baseTypesFor` returns
+                // undefined and the picker uses its own default. Must not throw — an exception
+                // inside a VTL <script> takes the whole custom field down with it.
+                const warn = jest.spyOn(console, 'warn').mockImplementation();
+
+                expect(() =>
+                    bridge.openBrowserModal({
+                        kinds: ['folder']
+                    } as unknown as DotBrowserOptions)
+                ).not.toThrow();
+
+                expect(openedConfig().allowedBaseTypes).toEqual(['DOTASSET', 'FILEASSET']);
+                expect(warn).toHaveBeenCalledTimes(1);
+
+                warn.mockRestore();
+            });
+
+            it('should not warn for a caller whose kinds are all supported', () => {
+                const warn = jest.spyOn(console, 'warn').mockImplementation();
+
+                bridge.openBrowserModal({ kinds: ['file', 'dotasset', 'page', 'link'] });
+
+                expect(warn).not.toHaveBeenCalled();
+
+                warn.mockRestore();
             });
 
             it.each([
@@ -1069,26 +1124,9 @@ describe('AngularFormBridge', () => {
                 );
             });
 
-            it('should report a folder with its path as the url', () => {
-                // A folder has no `url` — its path is what a custom field stores.
-                const onClose = jest.fn();
-                bridge.openBrowserModal({ kinds: ['folder'], onClose });
-                closeWith({
-                    type: 'folder',
-                    identifier: 'folder-1',
-                    inode: 'folder-inode',
-                    title: 'images',
-                    path: '/images/'
-                });
-
-                expect(onClose).toHaveBeenCalledWith({
-                    kind: 'folder',
-                    identifier: 'folder-1',
-                    inode: 'folder-inode',
-                    title: 'images',
-                    url: '/images/'
-                });
-            });
+            // Removed in #37366: "should report a folder with its path as the url". A folder can no
+            // longer be picked, so there is no selection to report. The path a custom field stores
+            // for a folder is now typed into the field, not returned by the picker.
 
             it('should report a menu link with its target as the url', () => {
                 const onClose = jest.fn();
@@ -1112,10 +1150,8 @@ describe('AngularFormBridge', () => {
             });
 
             it.each([
-                [
-                    'folder',
-                    { type: 'folder', identifier: 'f', inode: 'fi', title: 'f', path: '/f/' }
-                ],
+                // The `folder` case went with #37366 — a folder is no longer a selectable kind, so
+                // there is no folder selection whose shape could be wrong.
                 ['link', { extension: 'link', identifier: 'l', inode: 'li', title: 'l', url: '/l' }]
             ])('should not attach contentlet-only fields to a %s', (_kind, item) => {
                 // The whole point of the discriminated union: a consumer can never read a mimetype
diff --git a/core-web/libs/edit-content-bridge/src/lib/bridges/angular-form-bridge.ts b/core-web/libs/edit-content-bridge/src/lib/bridges/angular-form-bridge.ts
index 3fe28bb9a27f..f8f6cc6f23d4 100644
--- a/core-web/libs/edit-content-bridge/src/lib/bridges/angular-form-bridge.ts
+++ b/core-web/libs/edit-content-bridge/src/lib/bridges/angular-form-bridge.ts
@@ -618,6 +618,14 @@ export class AngularFormBridge implements FormBridge {
     }
 }
 
+/**
+ * Every kind the browse contract admits.
+ *
+ * Declared rather than derived from {@link KIND_BASE_TYPES}, which covers only the kinds that map
+ * onto a base type — `link` has none, so deriving would report it as unsupported.
+ */
+const SUPPORTED_KINDS: readonly DotBrowserItemKind[] = ['file', 'dotasset', 'page', 'link'];
+
 /** Base types a `kinds` list asks the selector to offer. */
 const KIND_BASE_TYPES: Partial<Record<DotBrowserItemKind, string>> = {
     file: 'FILEASSET',
@@ -651,6 +659,21 @@ function browseOptionsFor(options: DotBrowserOptions): DotAssetPickerBrowseOptio
     const kinds = options.kinds ?? [];
     const wantsLinks = kinds.includes('link');
 
+    // Callers are VTL string literals, so `kinds` can carry anything the type no longer admits —
+    // `'folder'` above all, which `file_browser_field_render_new.vtl` asked for until #37366, and
+    // which a third-party template still might. Warn and drop it: an
+    // exception here would take the whole custom field down, while the rest of the request is still
+    // satisfiable. Mirrors the mimetype conflict below.
+    const unsupported = kinds.filter((kind) => !SUPPORTED_KINDS.includes(kind));
+
+    if (unsupported.length) {
+        console.warn(
+            `DotCustomFieldApi.openBrowserModal: unsupported kind(s) ${unsupported.join(', ')} — ` +
+                'ignored. Folders in particular are navigation, not content: they appear only in ' +
+                'the picker sidebar tree and can never be listed or returned.'
+        );
+    }
+
     if (wantsLinks && options.mimeTypes?.length) {
         // The browse endpoint drops links whenever a mimetype filter is set, because a link has no
         // file metadata to match against. Surfaced rather than worked around: silently returning
@@ -662,7 +685,6 @@ function browseOptionsFor(options: DotBrowserOptions): DotAssetPickerBrowseOptio
     }
 
     return {
-        ...(kinds.includes('folder') ? { showFolders: true } : {}),
         ...(wantsLinks ? { showLinks: true } : {}),
         // Three states, expressed as the two flags the picker already understands.
         ...(options.status ? { showWorking: options.status !== 'live' } : {}),
@@ -673,12 +695,12 @@ function browseOptionsFor(options: DotBrowserOptions): DotAssetPickerBrowseOptio
     };
 }
 
-/** What the picker returned, in the terms the item itself reports. */
+/**
+ * What the picker returned, in the terms the item itself reports.
+ *
+ * No folder branch: the picker never lists a folder, so a row can never be one.
+ */
 function kindOf(item: Record<string, unknown>): DotBrowserItemKind {
-    if (item['type'] === 'folder') {
-        return 'folder';
-    }
-
     if (item['type'] === 'link' || item['extension'] === 'link') {
         return 'link';
     }
@@ -695,9 +717,9 @@ function kindOf(item: Record<string, unknown>): DotBrowserItemKind {
 /**
  * Maps a picked row onto the published selection shape.
  *
- * `url` is the one guarantee: a contentlet reports `url` (or `urlMap`), a folder its path, a link
- * its target. Contentlet-only fields are attached only for the kinds that actually have them, so a
- * consumer can never read a mimetype off a folder.
+ * `url` is the one guarantee: a contentlet reports `url` (or `urlMap`), a link its target.
+ * Contentlet-only fields are attached only for the kinds that actually have them, so a consumer can
+ * never read a mimetype off a menu link.
  */
 /**
  * A string, or nothing.
@@ -721,7 +743,7 @@ function toSelection(item: DotCMSContentlet | Record<string, unknown>): DotBrows
         url: String(row['url'] ?? row['urlMap'] ?? row['path'] ?? '')
     };
 
-    if (kind === 'folder' || kind === 'link') {
+    if (kind === 'link') {
         return base as DotBrowserSelection;
     }
 
diff --git a/core-web/libs/edit-content-bridge/src/lib/interfaces/asset-browser.interface.ts b/core-web/libs/edit-content-bridge/src/lib/interfaces/asset-browser.interface.ts
index 69693fd0b5b9..2b9dca483946 100644
--- a/core-web/libs/edit-content-bridge/src/lib/interfaces/asset-browser.interface.ts
+++ b/core-web/libs/edit-content-bridge/src/lib/interfaces/asset-browser.interface.ts
@@ -3,10 +3,14 @@
  * `DotCustomFieldApi.openBrowserModal()`.
  *
  * Replaces the previous `BrowserSelector*` shapes, which were named after the legacy
- * `DotBrowserSelectorComponent` and modelled every selection as a contentlet — so a folder or a
- * menu link had to carry a mimetype and a base type it does not have. That was safe to change
- * because the API is unshipped: the templates that call it (`*_new.vtl`) only render when the new
- * Edit Content is enabled, which is not the default.
+ * `DotBrowserSelectorComponent` and modelled every selection as a contentlet — so a menu link had
+ * to carry a mimetype and a base type it does not have. That was safe to change because the API is
+ * unshipped: the templates that call it (`*_new.vtl`) only render when the new Edit Content is
+ * enabled, which is not the default.
+ *
+ * The same unshipped status is why `'folder'` could be withdrawn outright in #37366 rather than
+ * deprecated: folders are navigation, reached through the picker's sidebar tree, and are neither
+ * listed nor returned.
  */
 
 /**
@@ -15,8 +19,12 @@
  * `file` and `dotasset` are kept apart rather than collapsed into one `asset` value because callers
  * genuinely distinguish them — the shipped file-browser template asks for file assets while
  * excluding dotAssets.
+ *
+ * `folder` is deliberately absent. Folders are navigation, not content: they appear only in the
+ * picker's sidebar tree, where selecting one changes what the list shows. No option lists a folder
+ * as a row and none returns one.
  */
-export type DotBrowserItemKind = 'file' | 'dotasset' | 'page' | 'folder' | 'link';
+export type DotBrowserItemKind = 'file' | 'dotasset' | 'page' | 'link';
 
 /**
  * What a caller asks the browser to show.
@@ -31,9 +39,20 @@ export interface DotBrowserOptions {
     /**
      * What may be listed and returned.
      *
-     * One list rather than five booleans: `showFiles`/`showPages`/`showFolders`/`showLinks`/
-     * `showDotAssets` could express combinations that mean nothing, and could not say "these kinds"
-     * without naming every other kind too.
+     * One list rather than a boolean per kind: `showFiles`/`showPages`/`showLinks`/`showDotAssets`
+     * could express combinations that mean nothing, and could not say "these kinds" without naming
+     * every other kind too.
+     *
+     * **Folders cannot be requested and cannot be returned.** They are navigation, not content:
+     * they appear only in the picker's sidebar tree, where selecting one changes what the list
+     * shows. To store a folder path in a field, type it — the picker will not hand you one.
+     *
+     * An entry that is not a {@link DotBrowserItemKind} is **ignored with a console warning**
+     * rather than throwing, since a caller is a VTL `<script>` where an exception would break the
+     * whole custom field. `'folder'` is the case this exists for: the shipped file-browser template
+     * (`file_browser_field_render_new.vtl`) asked for it before the kind was withdrawn, and a
+     * third-party template still might. The other shipped caller, `redirect_custom_field_new.vtl`,
+     * never did.
      *
      * @default ['file', 'dotasset']
      */
@@ -107,8 +126,7 @@ export interface DotBrowserSelectionBase {
      * The value a field stores. **Always non-empty.**
      *
      * The one field every shipped template actually reads, so an empty value is a defect rather
-     * than a degraded result: a contentlet's URL, a page's URL, a folder's path, or a link's
-     * target.
+     * than a degraded result: a contentlet's URL, a page's URL, or a link's target.
      */
     url: string;
 }
@@ -130,11 +148,6 @@ export interface DotBrowserPageSelection extends DotBrowserSelectionBase {
     contentType?: string;
 }
 
-/** A folder. `url` is its path. */
-export interface DotBrowserFolderSelection extends DotBrowserSelectionBase {
-    kind: 'folder';
-}
-
 /** A menu link. `url` is its target. */
 export interface DotBrowserLinkSelection extends DotBrowserSelectionBase {
     kind: 'link';
@@ -144,13 +157,12 @@ export interface DotBrowserLinkSelection extends DotBrowserSelectionBase {
  * What the editor picked.
  *
  * A union discriminated by `kind` rather than one flat shape with everything optional: the flat
- * shape let a consumer read a mimetype off a folder and get `undefined` with no indication that the
- * question was meaningless. Branch on `kind` and each variant offers exactly the fields it has.
+ * shape let a consumer read a mimetype off a menu link and get `undefined` with no indication that
+ * the question was meaningless. Branch on `kind` and each variant offers exactly the fields it has.
  */
 export type DotBrowserSelection =
     | DotBrowserAssetSelection
     | DotBrowserPageSelection
-    | DotBrowserFolderSelection
     | DotBrowserLinkSelection;
 
 /**
diff --git a/core-web/libs/edit-content-bridge/src/lib/interfaces/form-bridge.interface.ts b/core-web/libs/edit-content-bridge/src/lib/interfaces/form-bridge.interface.ts
index fb7c27e4d917..fe4ada335e79 100644
--- a/core-web/libs/edit-content-bridge/src/lib/interfaces/form-bridge.interface.ts
+++ b/core-web/libs/edit-content-bridge/src/lib/interfaces/form-bridge.interface.ts
@@ -49,8 +49,10 @@ export interface FormBridge {
     destroy(): void;
 
     /**
-     * Opens the asset browser so the user can pick content — an asset, a page, a folder or a menu
-     * link.
+     * Opens the asset browser so the user can pick content — an asset, a page or a menu link.
+     *
+     * Folders are **not** pickable: they are navigation, reached through the browser's sidebar
+     * tree, and the list panel carries content only.
      *
      * Only the Angular host opens anything: the legacy Dojo editor has never had this dialog, and
      * its bridge resolves `null` with a warning rather than pretending.
diff --git a/core-web/libs/ui/src/lib/components/dot-asset-picker/asset-picker-config.spec.ts b/core-web/libs/ui/src/lib/components/dot-asset-picker/asset-picker-config.spec.ts
index eda72f65ce87..83d6beee819e 100644
--- a/core-web/libs/ui/src/lib/components/dot-asset-picker/asset-picker-config.spec.ts
+++ b/core-web/libs/ui/src/lib/components/dot-asset-picker/asset-picker-config.spec.ts
@@ -225,8 +225,9 @@ describe('buildAssetPickerConfig', () => {
         });
 
         it('should carry the browse options through untouched', () => {
+            // No `showFolders` — the option left `DotAssetPickerBrowseOptions` in #37366. The
+            // picker's list carries content only, so there is nothing for a caller to opt into.
             const browse = {
-                showFolders: true,
                 showLinks: true,
                 showWorking: false,
                 showArchived: false,
@@ -238,6 +239,16 @@ describe('buildAssetPickerConfig', () => {
             expect(config.browse).toEqual(browse);
         });
 
+        it('should never carry a folder browse option', () => {
+            const config = buildAssetPickerConfig({
+                mode: 'browse',
+                site: SITE,
+                browse: { showLinks: true }
+            });
+
+            expect(config.browse).not.toHaveProperty('showFolders');
+        });
+
         it('should use the caller-supplied title', () => {
             // The picker renders its own header, so the title travels in the config rather than in
             // `DynamicDialogConfig.header`.
diff --git a/core-web/libs/ui/src/lib/components/dot-asset-picker/store/constants.ts b/core-web/libs/ui/src/lib/components/dot-asset-picker/store/constants.ts
index 9a226c1e9532..875d2b8dd4ac 100644
--- a/core-web/libs/ui/src/lib/components/dot-asset-picker/store/constants.ts
+++ b/core-web/libs/ui/src/lib/components/dot-asset-picker/store/constants.ts
@@ -14,8 +14,6 @@ export const DEFAULT_ASSET_PICKER_SORT: DotAssetPickerSort = {
 export const DEFAULT_ASSET_PICKER_PAGE: DotAssetPickerPage = {
     contentCursor: 0,
     hasMoreContent: true,
-    folderCursor: 0,
-    hasMoreFolders: true,
     linkCursor: 0,
     hasMoreLinks: true
 };
diff --git a/core-web/libs/ui/src/lib/components/dot-asset-picker/store/dot-asset-picker.store.spec.ts b/core-web/libs/ui/src/lib/components/dot-asset-picker/store/dot-asset-picker.store.spec.ts
index 535682c5d694..58bbda40343d 100644
--- a/core-web/libs/ui/src/lib/components/dot-asset-picker/store/dot-asset-picker.store.spec.ts
+++ b/core-web/libs/ui/src/lib/components/dot-asset-picker/store/dot-asset-picker.store.spec.ts
@@ -94,13 +94,17 @@ const IMAGE_FIELD_CONFIG: DotAssetPickerConfig = {
 };
 
 /**
- * What `openBrowserModal` hands the store: pages alongside assets, plus folders and menu links —
- * none of which any other entry point may ask for.
+ * What `openBrowserModal` hands the store: pages alongside assets, plus menu links — which no other
+ * entry point may ask for.
+ *
+ * No `showFolders`: the option no longer exists (#37366). The picker's list carries content only,
+ * and folders are reached through the sidebar tree, so there is no browse option a caller could set
+ * to put a folder in the list.
  */
 const BROWSE_CONFIG: DotAssetPickerConfig = {
     site: SITE,
     allowedBaseTypes: [DotCMSBaseTypesContentTypes.FILEASSET, DotCMSBaseTypesContentTypes.HTMLPAGE],
-    browse: { showFolders: true, showLinks: true }
+    browse: { showLinks: true }
 };
 
 describe('DotAssetPickerStore', () => {
@@ -245,12 +249,18 @@ describe('DotAssetPickerStore', () => {
         });
     });
 
-    // Formerly "showFolders invariant". It is no longer an invariant — `openBrowserModal` can ask
-    // for folders — but it remains the DEFAULT, and these are now the guards that prove no other
-    // entry point gained them.
-    describe('showFolders default (no browse options)', () => {
-        it('should be false with no filters applied', () => {
-            store.initPicker(FILE_FIELD_CONFIG);
+    // An invariant again (#37366). It briefly stopped being one when `openBrowserModal` could ask
+    // for folders; that capability is withdrawn, so `showFolders` is now false for EVERY entry
+    // point and there is no configuration that flips it. BROWSE_CONFIG is asserted alongside the
+    // field configs on purpose — it is the only caller that ever got folders, so a guard that
+    // omitted it would be a guard that cannot fail.
+    describe('showFolders invariant (every entry point)', () => {
+        it.each([
+            ['File field', FILE_FIELD_CONFIG],
+            ['Image field', IMAGE_FIELD_CONFIG],
+            ['browse (openBrowserModal)', BROWSE_CONFIG]
+        ])('should be false with no filters applied for the %s', (_name, config) => {
+            store.initPicker(config);
 
             expect(store.$request().showFolders).toBe(false);
         });
@@ -267,6 +277,18 @@ describe('DotAssetPickerStore', () => {
             expect(store.$request().showFolders).toBe(false);
         });
 
+        it('should be false with every filter applied for a browse caller', () => {
+            store.initPicker(BROWSE_CONFIG);
+            store.patchFilters({
+                title: 'logo',
+                contentType: ['fileAsset'],
+                baseType: ['FILEASSET'],
+                languageId: ['1', '2']
+            });
+
+            expect(store.$request().showFolders).toBe(false);
+        });
+
         it('should be false after filters are cleared', () => {
             store.initPicker(FILE_FIELD_CONFIG);
             store.patchFilters({ title: 'logo' });
@@ -275,10 +297,36 @@ describe('DotAssetPickerStore', () => {
             expect(store.$request().showFolders).toBe(false);
         });
 
-        it('should never advance the folder cursor', () => {
+        it('should send showFolders rather than omit it', () => {
+            // Load-bearing, and the reason this is asserted separately from the `false` checks:
+            // `showFolders` defaults to TRUE server-side (AbstractDriveRequestForm#showFolders), so
+            // an omitted key relists folders. `toBe(false)` alone would pass on `undefined` under a
+            // looser matcher, so assert the key's presence explicitly.
+            store.initPicker(BROWSE_CONFIG);
+
+            expect(store.$request()).toHaveProperty('showFolders', false);
+        });
+
+        it('should ignore a re-added showFolders browse option', () => {
+            // The one falsifiable guard at this layer, and the regression actually worth guarding:
+            // that no configuration can re-enable folder listing. The cast simulates a future
+            // change re-introducing the option that #37366 withdrew — the request must still send
+            // `false`, because the store hardcodes it rather than reading it from `browse`.
+            store.initPicker({
+                ...BROWSE_CONFIG,
+                browse: { showFolders: true }
+            } as unknown as DotAssetPickerConfig);
+
+            expect(store.$request().showFolders).toBe(false);
+        });
+
+        it('should not send a folder cursor at all', () => {
+            // Was "should never advance the folder cursor", asserting a pinned 0. The key is gone
+            // entirely now (#37366) — the endpoint defaults it to 0, and there is no folder stream
+            // for the picker to page.
             store.initPicker(FILE_FIELD_CONFIG);
 
-            expect(store.$request().folderCursor).toBe(0);
+            expect(store.$request()).not.toHaveProperty('folderCursor');
         });
     });
 
@@ -435,10 +483,14 @@ describe('DotAssetPickerStore', () => {
             nextLinkCursor: 3
         };
 
-        it('should ask for folders when the caller does', () => {
+        it('should never ask for folders, not even for a browse caller', () => {
+            // Inverted in #37366. #37207 shipped this as "should ask for folders when the caller
+            // does"; QA rejected the design (TC-001), so the capability is withdrawn rather than
+            // the implementation fixed. The list carries content only; the sidebar tree is where
+            // folders live.
             store.initPicker(BROWSE_CONFIG);
 
-            expect(store.$request().showFolders).toBe(true);
+            expect(store.$request().showFolders).toBe(false);
         });
 
         it('should ask for menu links when the caller does', () => {
@@ -465,61 +517,42 @@ describe('DotAssetPickerStore', () => {
             expect(store.$request().archived).toBe(true);
         });
 
-        describe('three-cursor paging', () => {
+        // Two cursors, not three (#37366). Contentlets and menu links still page independently;
+        // the folder stream is gone, so the request carries no folder cursor to resume.
+        describe('two-cursor paging', () => {
             beforeEach(() => {
                 contentDriveService.search.mockReturnValue(of(MORE_OF_EVERYTHING));
                 store.initPicker(BROWSE_CONFIG);
                 spectator.flushEffects();
             });
 
-            it('should no longer pin the folder cursor to zero', () => {
-                // The old invariant ("with showFolders: false the folder cursor never advances")
-                // held only because folders were never returned. Now they can be.
+            it('should not carry a folder cursor on any page', () => {
+                expect(store.$request()).not.toHaveProperty('folderCursor');
+
                 store.setPagination({ ...DEFAULT_ASSET_PICKER_PAGINATION, page: 2 });
 
-                expect(store.$request().folderCursor).toBe(5);
+                expect(store.$request()).not.toHaveProperty('folderCursor');
             });
 
             it('should resume every stream from where the previous page left off', () => {
-                // Contentlets, folders and links page independently — one cursor cannot describe a
-                // page of a mixed list.
+                // Contentlets and links page independently — one cursor cannot describe a page of a
+                // mixed list.
                 store.setPagination({ ...DEFAULT_ASSET_PICKER_PAGINATION, page: 2 });
 
                 expect(store.$request().contentCursor).toBe(20);
-                expect(store.$request().folderCursor).toBe(5);
                 expect(store.$request().linkCursor).toBe(3);
             });
 
             it('should start every stream at zero on page one', () => {
                 expect(store.$request().contentCursor).toBe(0);
-                expect(store.$request().folderCursor).toBe(0);
                 expect(store.$request().linkCursor).toBe(0);
             });
         });
 
         describe('exhausted streams', () => {
-            it('should stop asking for folders once there are no more', () => {
-                // The endpoint's own contract: when hasMoreFolders comes back false, the next page
-                // should skip the folder query entirely rather than pay for it again.
-                contentDriveService.search.mockReturnValue(
-                    of({
-                        ...MORE_OF_EVERYTHING,
-                        hasMoreFolders: false,
-                        nextFolderCursor: 5
-                    })
-                );
-                store.initPicker(BROWSE_CONFIG);
-                spectator.flushEffects();
-
-                // Asserted on BOTH pages on purpose. `showFolders` starts out hardcoded `false`, so
-                // checking only page 2 would pass today for entirely the wrong reason — a guard
-                // that cannot fail is not a guard.
-                expect(store.$request().showFolders).toBe(true);
-
-                store.setPagination({ ...DEFAULT_ASSET_PICKER_PAGINATION, page: 2 });
-
-                expect(store.$request().showFolders).toBe(false);
-            });
+            // "should stop asking for folders once there are no more" was removed in #37366: there
+            // is no folder stream to exhaust. `showFolders` is a hardcoded `false` on every page,
+            // which the `showFolders invariant` suite above covers.
 
             it('should stop asking for links once there are no more', () => {
                 contentDriveService.search.mockReturnValue(
@@ -555,7 +588,7 @@ describe('DotAssetPickerStore', () => {
         };
 
         beforeEach(() => {
-            siteService.getCurrentSite = jest.fn().mockReturnValue(of(SITE));
+            siteService.getCurrentSite.mockReturnValue(of(SITE));
         });
 
         it('should not ask the server when the caller already supplied a site', () => {
@@ -578,7 +611,7 @@ describe('DotAssetPickerStore', () => {
         it('should not search until a site is known', () => {
             // `$isBrowsable` already guards on `browsingSite`, so the request simply waits rather
             // than firing against an undefined path.
-            siteService.getCurrentSite = jest.fn().mockReturnValue(NEVER);
+            siteService.getCurrentSite.mockReturnValue(NEVER);
 
             store.initPicker(NO_SITE_CONFIG);
             spectator.flushEffects();
@@ -590,9 +623,7 @@ describe('DotAssetPickerStore', () => {
         it('should still open when the lookup fails, leaving the site tree unselected', () => {
             // Opening on the site tree with nothing chosen beats not opening at all — the sidebar
             // lists every site the user can browse, so they can pick one.
-            siteService.getCurrentSite = jest
-                .fn()
-                .mockReturnValue(throwError(() => new Error('offline')));
+            siteService.getCurrentSite.mockReturnValue(throwError(() => new Error('offline')));
 
             store.initPicker(NO_SITE_CONFIG);
             spectator.flushEffects();
@@ -614,12 +645,15 @@ describe('DotAssetPickerStore', () => {
         });
     });
 
-    describe('paginator row count across three streams', () => {
-        it('should keep Next reachable while folders remain, even with content exhausted', () => {
-            // Reported in review of #37273. The paginator total looked only at `hasMoreContent`, so
-            // a page whose content stream ended while folders kept going reported the rows already
-            // on screen as the grand total. PrimeNG then sees `first + rows >= totalRecords`,
-            // disables Next, and the remaining folders become unreachable.
+    // Two streams since #37366, and the invariant is unchanged: claim one page beyond while ANY
+    // surviving stream reports more, so a stream that outlives the others stays reachable. The
+    // folder case that #37273's review originally caught is gone with the folder stream itself;
+    // the menu-link case below is the same bug shape and is what now guards it.
+    describe('paginator row count across two streams', () => {
+        it('should ignore hasMoreFolders even if the endpoint reports it', () => {
+            // The response type still carries the folder pair, and a `showFolders: false` request
+            // always gets `hasMoreFolders: false` — but a server that reported `true` must not
+            // inflate the total for rows the picker never asked for and cannot show.
             contentDriveService.search.mockReturnValue(
                 of({
                     ...EMPTY_RESPONSE,
@@ -633,8 +667,7 @@ describe('DotAssetPickerStore', () => {
             store.initPicker(BROWSE_CONFIG);
             spectator.flushEffects();
 
-            // One page beyond what is on screen, so the arrow stays live.
-            expect(store.$totalRecords()).toBeGreaterThan(15);
+            expect(store.$totalRecords()).toBe(15);
         });
 
         it('should keep Next reachable while menu links remain', () => {
@@ -686,7 +719,9 @@ describe('DotAssetPickerStore', () => {
         it.each([
             ['File field', FILE_FIELD_CONFIG],
             ['Image field', IMAGE_FIELD_CONFIG]
-        ])('should keep the folder cursor pinned for the %s', (_name, config) => {
+        ])('should send no folder cursor for the %s, on any page', (_name, config) => {
+            // Was "should keep the folder cursor pinned", asserting 0. The key is absent entirely
+            // now (#37366) — asserted on page 2 as well, since paging is what used to advance it.
             contentDriveService.search.mockReturnValue(
                 of({ ...EMPTY_RESPONSE, hasMoreContent: true, nextContentCursor: 20 })
             );
@@ -695,7 +730,7 @@ describe('DotAssetPickerStore', () => {
 
             store.setPagination({ ...DEFAULT_ASSET_PICKER_PAGINATION, page: 2 });
 
-            expect(store.$request().folderCursor).toBe(0);
+            expect(store.$request()).not.toHaveProperty('folderCursor');
         });
 
         it('should not request archived content', () => {
diff --git a/core-web/libs/ui/src/lib/components/dot-asset-picker/store/features/with-asset-browse.feature.ts b/core-web/libs/ui/src/lib/components/dot-asset-picker/store/features/with-asset-browse.feature.ts
index a7ab53b762bf..58bd1f57e9ff 100644
--- a/core-web/libs/ui/src/lib/components/dot-asset-picker/store/features/with-asset-browse.feature.ts
+++ b/core-web/libs/ui/src/lib/components/dot-asset-picker/store/features/with-asset-browse.feature.ts
@@ -49,8 +49,11 @@ const initialState: DotAssetPickerBrowseState = {
  *
  * Mirrors the shape of Content Drive's request builder, minus everything the picker has no use for
  * (workflow filters, user-searchable fields) and plus the two things it needs: a silent mimetype
- * restriction, and folders and menu links switched off unless the caller opts in through
+ * restriction, and menu links switched off unless the caller opts in through
  * `DotAssetPickerBrowseOptions`.
+ *
+ * Folders are switched off unconditionally — the list carries content only, and the sidebar tree is
+ * where folders are navigated.
  */
 export function withAssetBrowse() {
     return signalStoreFeature(
@@ -71,19 +74,21 @@ export function withAssetBrowse() {
              * page beyond to keep the arrow live; once none of them do, the page is the last one and
              * the exact total is knowable. Mirrors Content Drive's `$totalItems`.
              *
-             * All three streams matter, not just content: contentlets, folders and menu links page
-             * independently, so a page whose content ran out while folders kept going would
-             * otherwise report the rows already on screen as the grand total. PrimeNG then sees
-             * `first + rows >= totalRecords`, disables Next, and those folders become unreachable.
+             * Both streams matter, not just content: contentlets and menu links page independently,
+             * so a page whose content ran out while links kept going would otherwise report the rows
+             * already on screen as the grand total. PrimeNG then sees
+             * `first + rows >= totalRecords`, disables Next, and those links become unreachable.
+             *
+             * The response also carries `hasMoreFolders`, and it is deliberately NOT read: the
+             * picker never asks for folders, so folder rows can never be on screen and claiming a
+             * page for them would offer a page the user cannot fill.
              */
             $totalRecords: computed(() => {
                 const { page, limit } = pagination();
                 // `pages[N]` is the bookmark written AFTER loading page N, so its `hasMore*` flags
                 // answer "is there anything past what is on screen".
                 const bookmark = pages()[page];
-                const hasMore = bookmark
-                    ? bookmark.hasMoreContent || bookmark.hasMoreFolders || bookmark.hasMoreLinks
-                    : false;
+                const hasMore = bookmark ? bookmark.hasMoreContent || bookmark.hasMoreLinks : false;
 
                 return hasMore ? limit * (page + 1) : limit * (page - 1) + items().length;
             }),
@@ -126,8 +131,6 @@ export function withAssetBrowse() {
 
                     // Once a stream is exhausted the endpoint asks us to switch it off, so later
                     // pages stop paying for a query with nothing left to return.
-                    const showFolders =
-                        Boolean(browse?.showFolders) && (bookmark?.hasMoreFolders ?? true);
                     const showLinks =
                         Boolean(browse?.showLinks) && (bookmark?.hasMoreLinks ?? true);
 
@@ -145,10 +148,9 @@ export function withAssetBrowse() {
                             ? pickerConfig.mimeTypes
                             : undefined,
                         contentCursor: bookmark?.contentCursor ?? 0,
-                        // No longer pinned. It stays 0 while folders are switched off — which is
-                        // every entry point but `openBrowserModal` — but a caller that asks for
-                        // folders pages through them like any other stream.
-                        folderCursor: bookmark?.folderCursor ?? 0,
+                        // No `folderCursor`: with `showFolders: false` the endpoint runs no folder
+                        // query, so there is nothing to resume from. Its server-side default is 0,
+                        // which is what the key would have carried anyway.
                         ...(browse?.showLinks
                             ? { showLinks, linkCursor: bookmark?.linkCursor ?? 0 }
                             : {}),
@@ -163,10 +165,14 @@ export function withAssetBrowse() {
                         // wants, so the key is omitted unless a caller explicitly asked for
                         // live-only.
                         ...(browse?.showWorking === false ? { live: true } : {}),
-                        // Default, not an invariant any more: the picker's list is for assets, and
-                        // folders are navigated through the sidebar tree — unless the caller
-                        // explicitly asked for them.
-                        showFolders
+                        // An invariant, and hardcoded rather than derived from `browse` so that no
+                        // configuration can reach it: the picker's list carries content only, and
+                        // folders are navigated through the sidebar tree.
+                        //
+                        // Sent, never omitted. The endpoint defaults `showFolders` to TRUE
+                        // (AbstractDriveRequestForm#showFolders), so dropping the key would list
+                        // folders again — silently, since nothing here would look wrong.
+                        showFolders: false
                     };
                 },
                 // Structural dedupe so a no-op recompute doesn't refire the search.
@@ -208,14 +214,18 @@ export function withAssetBrowse() {
                                 pages[page] = {
                                     contentCursor: response.nextContentCursor,
                                     hasMoreContent: response.hasMoreContent,
-                                    folderCursor: response.nextFolderCursor,
-                                    hasMoreFolders: response.hasMoreFolders,
-                                    // Only a `showLinks` response carries these, and unlike the
-                                    // folder pair they are both optional — so a response could
-                                    // report more links without saying where to resume. Falling
-                                    // back to the cursor we sent keeps the stream from rewinding to
-                                    // 0 and re-serving links that were already shown; once there
-                                    // are no more, 0 is correct because nothing will read it again.
+                                    // `nextFolderCursor` / `hasMoreFolders` are deliberately not
+                                    // bookmarked: with `showFolders: false` the endpoint leaves
+                                    // them at their initial values, so recording them would store
+                                    // two constants and let a stream the picker never asked for
+                                    // influence its paging.
+                                    //
+                                    // Only a `showLinks` response carries the link pair, and both
+                                    // of its fields are optional — so a response could report more
+                                    // links without saying where to resume. Falling back to the
+                                    // cursor we sent keeps the stream from rewinding to 0 and
+                                    // re-serving links that were already shown; once there are no
+                                    // more, 0 is correct because nothing will read it again.
                                     linkCursor: response.hasMoreLinks
                                         ? (response.nextLinkCursor ?? sent?.linkCursor ?? 0)
                                         : 0,
diff --git a/core-web/libs/ui/src/lib/components/dot-asset-picker/store/models.ts b/core-web/libs/ui/src/lib/components/dot-asset-picker/store/models.ts
index b1a572e783d3..584bcdfa784f 100644
--- a/core-web/libs/ui/src/lib/components/dot-asset-picker/store/models.ts
+++ b/core-web/libs/ui/src/lib/components/dot-asset-picker/store/models.ts
@@ -15,22 +15,18 @@ import { ComponentStatus, DotContentDriveBrowseItem, TreeNodeItem } from '@dotcm
 /**
  * The browse capabilities a host can opt into.
  *
- * Absent, the picker behaves exactly as it always has: assets only, no folders, no links, not
- * archived. That is the point of nesting these rather than flattening six flags onto
- * {@link DotAssetPickerConfig} — a File-field change cannot set `showFolders` by accident, and the
- * opt-in is visible at every call site.
+ * Absent, the picker behaves exactly as it always has: assets only, no links, not archived. That is
+ * the point of nesting these rather than flattening the flags onto {@link DotAssetPickerConfig} — a
+ * File-field change cannot set `showLinks` by accident, and the opt-in is visible at every call
+ * site.
+ *
+ * **Folders are not among the capabilities**, by design: the picker's list carries content only and
+ * folders are reached through the sidebar tree, so there is no flag a host could set to put one in
+ * the list.
  *
  * Only the `browse` entry point (`openBrowserModal`) sets this today.
  */
 export interface DotAssetPickerBrowseOptions {
-    /**
-     * List folders alongside assets, and allow one to be returned.
-     *
-     * Folders are still navigated through the sidebar tree; this is about what the *list* may
-     * contain and return.
-     */
-    showFolders?: boolean;
-
     /**
      * List menu links alongside assets, and allow one to be returned.
      *
@@ -176,17 +172,18 @@ export interface DotAssetPickerPagination {
  * Cursor bookmark for one visited page.
  *
  * The drive API pages by cursor, not offset, so reaching page N means having walked pages 1..N-1.
- * Contentlets, folders and menu links each page independently, so a bookmark has to record all
- * three cursors — one cursor cannot reconstruct a page of a mixed list.
+ * Contentlets and menu links each page independently, so a bookmark has to record both cursors —
+ * one cursor cannot reconstruct a page of a mixed list.
+ *
+ * No folder cursor: the picker never asks for folders, so the endpoint's folder stream is always
+ * empty for it and its cursor would be a permanent 0.
  *
- * The folder and link cursors stay at 0 while their stream is switched off, which is the case for
- * every entry point that does not opt into {@link DotAssetPickerBrowseOptions}.
+ * The link cursor stays at 0 while its stream is switched off, which is the case for every entry
+ * point that does not opt into {@link DotAssetPickerBrowseOptions}.
  */
 export interface DotAssetPickerPage {
     contentCursor: number;
     hasMoreContent: boolean;
-    folderCursor: number;
-    hasMoreFolders: boolean;
     linkCursor: number;
     hasMoreLinks: boolean;
 }
@@ -298,9 +295,11 @@ export interface DotAssetPickerSelectionState {
     /**
      * The single item the editor has picked, or `null`.
      *
-     * Widened from `DotCMSContentlet` because the list can now return folders and menu links, which
-     * are not contentlets. What follows from that: only a contentlet can be re-fetched with its
-     * full content before the picker closes — see `DotAssetPickerComponent.confirm`.
+     * Widened from `DotCMSContentlet` because the list can return menu links, which are not
+     * contentlets. What follows from that: only a contentlet can be re-fetched with its full
+     * content before the picker closes — see `DotAssetPickerComponent.confirm`.
+     *
+     * Folders are no longer part of the reason: the list never carries one.
      */
     selectedAsset: DotContentDriveBrowseItem | null;
 }
diff --git a/dotCMS/src/main/webapp/WEB-INF/velocity/static/content/file_browser_field_render_new.vtl b/dotCMS/src/main/webapp/WEB-INF/velocity/static/content/file_browser_field_render_new.vtl
index ed9a6983430d..fba85c94f624 100644
--- a/dotCMS/src/main/webapp/WEB-INF/velocity/static/content/file_browser_field_render_new.vtl
+++ b/dotCMS/src/main/webapp/WEB-INF/velocity/static/content/file_browser_field_render_new.vtl
@@ -19,8 +19,8 @@
     // Handle browse button click
     browseButton.addEventListener("click", () => {
       DotCustomFieldApi.openBrowserModal({
-        title: "$text.get('Select-a-file')",
-        kinds: ["file", "page", "folder"],
+        title: "Select a page",
+        kinds: ["page"],
         status: "live",
         sort: { field: "modDate", direction: "desc" },
         onClose: (selection) => {
diff --git a/specs/37366-assetpicker-list-content-only/contracts/README.md b/specs/37366-assetpicker-list-content-only/contracts/README.md
new file mode 100644
index 000000000000..4c557fcb992b
--- /dev/null
+++ b/specs/37366-assetpicker-list-content-only/contracts/README.md
@@ -0,0 +1,14 @@
+# Contracts — #37366
+
+This fix changes **no REST contract**. No JAX-RS resource, no `@Schema`, no `openapi.yaml`
+regeneration. What changes is the **VTL-facing browse API** that custom-field templates call, plus
+the shape of the request the Asset Picker sends to an unchanged endpoint.
+
+| File | What it pins down |
+|---|---|
+| [`openbrowsermodal-public-api.md`](./openbrowsermodal-public-api.md) | `DotCustomFieldApi.openBrowserModal()` after the change — the kinds it accepts, what it returns, and what a stale `kinds: ['folder']` caller gets |
+| [`drive-search-request-delta.md`](./drive-search-request-delta.md) | The exact `/api/v1/drive/search` request body the picker sends, before and after. The endpoint is not changed — only what the picker asks for |
+
+Both are deltas against the contract #37207 shipped
+([`specs/37207-openbrowsermodal-assetpicker/contracts/`](../../37207-openbrowsermodal-assetpicker/contracts/)),
+which is the baseline to read them against.
diff --git a/specs/37366-assetpicker-list-content-only/contracts/drive-search-request-delta.md b/specs/37366-assetpicker-list-content-only/contracts/drive-search-request-delta.md
new file mode 100644
index 000000000000..eb87f7339650
--- /dev/null
+++ b/specs/37366-assetpicker-list-content-only/contracts/drive-search-request-delta.md
@@ -0,0 +1,100 @@
+# Contract: what the Asset Picker sends to `POST /api/v1/drive/search`
+
+**The endpoint is not changed by #37366.** No Java, no `@Schema`, no `openapi.yaml`. This file pins
+down the **client side** of the contract — the request body the picker builds — because that is where
+the fix is enforced.
+
+Server-side reference: `AbstractDriveRequestForm.java` (request shape),
+`ContentDriveHelper.java` (mapping to `BrowserQuery`), `BrowserAPIImpl.java:1690-1790` (paging).
+Behavior is DB-first per
+[ADR-0018](https://github.com/dotCMS/platform-adrs/blob/main/decisions/0018-database-first-content-drive-search-with-index-deferred-text-filtering.md).
+
+---
+
+## The rule that governs this delta
+
+```java
+// AbstractDriveRequestForm.java:395-397
+@JsonProperty("showFolders")
+@Value.Default
+default boolean showFolders(){return true; }
+```
+
+**`showFolders` defaults to `true`.** A request that omits it gets folders. So the picker must send
+an explicit `false` — omitting the key is not equivalent and would silently reintroduce the bug this
+fix removes. `ContentDriveResource.java:62` states the endpoint "honours `showFolders` as sent", and
+`ContentDriveHelper.java:234-236` refuses to override an explicit value.
+
+`folderCursor`, by contrast, defaults to `0` (`AbstractDriveRequestForm.java:480-482`), so that key
+*is* safe to drop.
+
+---
+
+## Request delta
+
+Built in `with-asset-browse.feature.ts:118-172`.
+
+| Key | Before — non-browse callers (File, Image, video, audio) | Before — browse caller asking for folders | **After — every caller** |
+|---|---|---|---|
+| `showFolders` | `false` (computed: `Boolean(undefined) && …`) | `true`, until `hasMoreFolders` came back `false` | **`false`, always, explicitly** |
+| `folderCursor` | `0` (pinned) | advanced from the page bookmark (`5`, `10`, …) | **key omitted** |
+| `showLinks` | omitted | `true` while links remain | unchanged |
+| `linkCursor` | omitted | advanced from the bookmark | unchanged |
+| `contentCursor` | advanced | advanced | unchanged |
+| `assetPath`, `includeSystemHost`, `filters`, `language`, `contentTypes`, `baseTypes`, `mimeTypes`, `maxResults`, `sortBy`, `archived`, `live` | — | — | unchanged |
+
+### After the change — the browse request for `forwardTo`
+
+```jsonc
+{
+  "assetPath": "//demo.dotcms.com/",
+  "includeSystemHost": true,
+  "filters": { "text": "", "filterFolders": true },
+  "baseTypes": ["FILEASSET", "HTMLPAGE"],   // from kinds: ["file", "page"]
+  "contentCursor": 0,
+  "maxResults": 20,
+  "sortBy": "modDate:desc",
+  "archived": false,
+  "live": true,                              // from status: "live"
+  "showFolders": false                       // explicit — the server default is true
+}
+```
+
+`showLinks` / `linkCursor` appear only for a caller that asks for `'link'` — which `forwardTo` does
+not, and `redirecturl` does.
+
+---
+
+## Response fields the picker stops reading
+
+The response shape is unchanged; the picker simply ignores two fields.
+
+| Field | Value when `showFolders: false` | Why it can be ignored |
+|---|---|---|
+| `hasMoreFolders` | always `false` | initialised `false` and only reassigned inside `if (browserQuery.showFolders)` — `BrowserAPIImpl.java:1705-1730` |
+| `nextFolderCursor` | echoes the `folderCursor` sent, i.e. `0` | same block |
+
+Consequences in the picker:
+
+- `DotAssetPickerPage` drops `folderCursor` / `hasMoreFolders` (see [data-model.md §4](../data-model.md#4-dotassetpickerpage--the-per-page-cursor-bookmark)).
+- `$totalRecords` unions `hasMoreContent || hasMoreLinks` only — two streams, same invariant
+  (see [§6](../data-model.md#6-derived-state--totalrecords)).
+
+`response.list` now contains contentlets only, which is the whole point: `items` never receives a
+folder, so `DotFolderListViewComponent` never renders a folder row — without the component changing
+at all.
+
+---
+
+## Side effect worth knowing (not a goal)
+
+The server spends the page budget on folders **before** content:
+
+```java
+// BrowserAPIImpl.java:1725
+maxResults -= folderCount;
+```
+
+With folders off, a page returns up to the full `maxResults` in contentlets instead of fewer. Pages
+will therefore look *fuller* after the fix. This is an improvement, not a regression — but it means
+row counts in any manual comparison against the pre-fix behavior will legitimately differ.
diff --git a/specs/37366-assetpicker-list-content-only/contracts/openbrowsermodal-public-api.md b/specs/37366-assetpicker-list-content-only/contracts/openbrowsermodal-public-api.md
new file mode 100644
index 000000000000..312131785fda
--- /dev/null
+++ b/specs/37366-assetpicker-list-content-only/contracts/openbrowsermodal-public-api.md
@@ -0,0 +1,93 @@
+# Contract: `DotCustomFieldApi.openBrowserModal()` after #37366
+
+**Surface**: JavaScript, called from a custom-field VTL template's `<script>` block.
+**Reachable only** when the new Edit Content experience is enabled
+(`$structures.isNewEditModeEnabled()`), which is not the default — so this contract is **unshipped**
+and may narrow without a deprecation cycle (`asset-browser.interface.ts:1-11`).
+
+**Baseline**: the contract as shipped by #37207 /
+[`specs/37207-openbrowsermodal-assetpicker/contracts/openbrowsermodal-public-api.md`](../../37207-openbrowsermodal-assetpicker/contracts/openbrowsermodal-public-api.md).
+Only the deltas are spelled out here.
+
+---
+
+## Delta 1 — `kinds` loses `'folder'`
+
+```ts
+// before
+type DotBrowserItemKind = 'file' | 'dotasset' | 'page' | 'folder' | 'link';
+// after
+type DotBrowserItemKind = 'file' | 'dotasset' | 'page' | 'link';
+```
+
+| Value | Lists | Returnable |
+|---|---|---|
+| `'file'` | file assets | ✅ |
+| `'dotasset'` | dotAssets | ✅ |
+| `'page'` | HTML pages | ✅ |
+| `'link'` | menu links | ✅ |
+| ~~`'folder'`~~ | **never listed** | **never returned** |
+
+**Folders are navigation, not content.** They appear only in the picker's sidebar tree, where
+selecting one changes what the list shows. There is no option that makes a folder appear as a list
+row, and no option that makes a folder the picked value.
+
+`@default` is unchanged: `['file', 'dotasset']`.
+
+---
+
+## Delta 2 — the returned selection union loses its folder member
+
+```ts
+// after
+type DotBrowserSelection =
+    | DotBrowserAssetSelection   // kind: 'file' | 'dotasset'
+    | DotBrowserPageSelection    // kind: 'page'
+    | DotBrowserLinkSelection;   // kind: 'link'
+```
+
+`onClose(selection)` can no longer be called with `kind: 'folder'`. A consumer that branches on
+`kind` needs no change — the branch simply becomes unreachable. A consumer that only reads
+`selection.url` (both shipped templates do) is entirely unaffected.
+
+---
+
+## Delta 3 — behavior for a caller that still asks for `'folder'`
+
+TypeScript cannot police a value arriving from a VTL string literal, so this is specified as
+**runtime behavior** (the point of spec AC-008):
+
+| Caller passes | Result |
+|---|---|
+| `kinds: ['file', 'page', 'folder']` | Behaves as `['file', 'page']`. No folder rows. **No error.** One `console.warn` naming `folder` as unsupported |
+| `kinds: ['folder']` | No kind maps to a base type, so the picker falls back to its asset-only default (`baseTypesFor` returns `undefined` — `angular-form-bridge.ts:635-642`). One `console.warn` |
+
+The warning follows the idiom already in `browseOptionsFor` for the `link` + `mimeTypes` conflict
+(`angular-form-bridge.ts:654-662`): surface the mismatch rather than silently return fewer kinds
+than asked for. It never throws — an exception inside a VTL `<script>` would break the whole custom
+field.
+
+---
+
+## Unchanged
+
+Everything else in the contract stands exactly as #37207 shipped it:
+
+- `title`, `status` (`'live' | 'working' | 'archived'`), `path`, `mimeTypes`, `sort`, `onClose`
+- `DotBrowserController` and its `close()`
+- The `link` + `mimeTypes` warning
+- The legacy Dojo bridge still resolves `null` with a warning rather than opening anything
+- Menu-link listing and paging (#37112)
+
+---
+
+## Shipped callers
+
+| Template | Field | `kinds` before | `kinds` after |
+|---|---|---|---|
+| `file_browser_field_render_new.vtl:23` | Vanity URL `forwardTo` | `["file","page","folder"]` | **`["file","page"]`** |
+| `redirect_custom_field_new.vtl:57` | HTML page `redirecturl` | `["page","link"]` | unchanged |
+
+A folder path can still be **set** on `forwardTo` — the field keeps its free-text input
+(`file_browser_field_render_new.vtl:38-44`), which writes straight to the field value. It can no
+longer be **picked** from the browser.
diff --git a/specs/37366-assetpicker-list-content-only/data-model.md b/specs/37366-assetpicker-list-content-only/data-model.md
new file mode 100644
index 000000000000..1bd1ed6c42d9
--- /dev/null
+++ b/specs/37366-assetpicker-list-content-only/data-model.md
@@ -0,0 +1,167 @@
+# Phase 1 Data Model: type and state deltas
+
+**Feature**: [`spec.md`](./spec.md) · **Plan**: [`plan.md`](./plan.md) · **Research**: [`research.md`](./research.md)
+
+There is no persistent data model here — nothing is stored, migrated or serialized. The "entities"
+this fix changes are **TypeScript types and store state**. Each delta below is the authoritative
+target shape; `/speckit-tasks` turns them into tasks.
+
+---
+
+## 1. `DotBrowserItemKind` — the VTL-facing kind vocabulary
+
+`core-web/libs/edit-content-bridge/src/lib/interfaces/asset-browser.interface.ts:19`
+
+| Before | After |
+|---|---|
+| `'file' \| 'dotasset' \| 'page' \| 'folder' \| 'link'` | `'file' \| 'dotasset' \| 'page' \| 'link'` |
+
+**Validation rule** (new, enforced at runtime not by the type — see [R3](./research.md#r3)): a
+`kinds` entry that is not one of the four is ignored, and the bridge warns once naming it. A VTL
+caller can pass any string, so the type alone guarantees nothing.
+
+**Consumers**: `angular-form-bridge.ts` only (`:18`, `:622`, `:635`, `:677`). Not in the lib's public
+index — see [R8](./research.md#r8).
+
+---
+
+## 2. `DotBrowserSelection` — what a pick resolves to
+
+`asset-browser.interface.ts:132-135` and the union that follows
+
+| Member | Fate |
+|---|---|
+| `DotBrowserAssetSelection` (file / dotasset) | unchanged |
+| `DotBrowserPageSelection` | unchanged |
+| `DotBrowserLinkSelection` | unchanged |
+| **`DotBrowserFolderSelection`** | **removed**, along with its arm of the `DotBrowserSelection` union (`:150-154`) |
+
+**State transition removed**: there is no longer a path by which `onClose(selection)` receives
+`kind: 'folder'`. The mapper's `kind === 'folder' || kind === 'link'` branch
+(`angular-form-bridge.ts:724`) narrows to links only, and `kindOf`'s `item['type'] === 'folder'`
+branch (`:678-679`) is deleted — nothing in `items` can be a folder any more, so the branch is
+unreachable rather than merely unused.
+
+---
+
+## 3. `DotAssetPickerBrowseOptions` — the picker's browse opt-ins
+
+`core-web/libs/ui/src/lib/components/dot-asset-picker/store/models.ts:26-32`
+
+| Field | Fate |
+|---|---|
+| **`showFolders?: boolean`** | **removed** |
+| `showLinks?: boolean` | unchanged — the link stream stays (#37112) |
+| `showWorking?: boolean` | unchanged |
+| `showArchived?: boolean` | unchanged |
+| `sortByDesc?: boolean` | unchanged |
+| `sortField?: string` | unchanged |
+
+The interface's own TSDoc still says *"Absent, the picker behaves exactly as it always has: assets
+only, no folders, no links…"* — that sentence stays true and now describes **every** case, not just
+the default. The `showFolders` field doc (which currently reads "List folders alongside assets, and
+allow one to be returned") goes with the field.
+
+---
+
+## 4. `DotAssetPickerPage` — the per-page cursor bookmark
+
+`store/models.ts:185-192`, seeded by `store/constants.ts:13-21`
+
+| Field | Fate |
+|---|---|
+| `contentCursor: number` / `hasMoreContent: boolean` | unchanged |
+| **`folderCursor: number`** / **`hasMoreFolders: boolean`** | **removed** ([R4](./research.md#r4), [R5](./research.md#r5)) |
+| `linkCursor: number` / `hasMoreLinks: boolean` | unchanged |
+
+`DEFAULT_ASSET_PICKER_PAGE` loses the same two fields. Safe because a `showFolders: false` response
+always returns `hasMoreFolders: false` and echoes back the `folderCursor` it was sent — both would
+be constants (`BrowserAPIImpl.java:1705-1730`).
+
+---
+
+## 5. The outgoing request — `DotContentDriveSearchRequest`
+
+Built in `store/features/with-asset-browse.feature.ts:118-172`. The **type** is shared with Content
+Drive (`libs/dotcms-models/src/lib/dot-content-drive.model.ts:246, 298`) and is **not** changed —
+Content Drive still uses both keys.
+
+| Key | Before (picker) | After (picker) |
+|---|---|---|
+| `showFolders` | `Boolean(browse?.showFolders) && (bookmark?.hasMoreFolders ?? true)` (`:129-130`) | constant **`false`** — never omitted ([R2](./research.md#r2)) |
+| `folderCursor` | `bookmark?.folderCursor ?? 0` (`:151`) | **key not sent** (server default `0`) |
+| `showLinks` / `linkCursor` | conditional on `browse?.showLinks` | unchanged |
+| `contentCursor` | `bookmark?.contentCursor ?? 0` | unchanged |
+| everything else | — | unchanged |
+
+> ⚠️ `showFolders` must be **present and `false`**. The server default is `true`
+> (`AbstractDriveRequestForm.java:395-397`); omitting the key relists folders.
+
+---
+
+## 6. Derived state — `$totalRecords`
+
+`with-asset-browse.feature.ts:79-89`
+
+| Before | After |
+|---|---|
+| `hasMoreContent \|\| hasMoreFolders \|\| hasMoreLinks` (`:85`) | `hasMoreContent \|\| hasMoreLinks` |
+
+The *purpose* is unchanged and must stay intact: claim one page beyond while **any** surviving
+stream reports more, so a stream that outlives the others stays reachable. Two streams instead of
+three; the invariant is the same. This is the #37207 TC-007 surface.
+
+---
+
+## 7. The shipped template
+
+`dotCMS/src/main/webapp/WEB-INF/velocity/static/content/file_browser_field_render_new.vtl:23`
+
+| Before | After |
+|---|---|
+| `kinds: ["file", "page", "folder"],` | `kinds: ["file", "page"],` |
+
+Nothing else in the template changes — notably not the free-text `#vlUri` input (`:38-44`), which is
+how a folder path can still be *set*.
+
+`redirect_custom_field_new.vtl:57` already reads `kinds: ["page", "link"]` and is **not edited**.
+
+---
+
+## 8. Documentation surface (AC-008)
+
+Not types, but part of the contract and therefore listed here so nothing is missed:
+
+| Location | What must change |
+|---|---|
+| `form-bridge.interface.ts:52-53` | `openBrowserModal` summary — "an asset, a page, a folder or a menu link" drops the folder |
+| `asset-browser.interface.ts` `kinds` doc (`:31-40`) | state that folders are navigation-only, cannot be requested and cannot be returned |
+| `asset-browser.interface.ts:6-9` (file header) | mentions "a folder or a menu link" as the shapes the old API mis-modelled — keep the history, but stop implying folders are still returnable |
+| `store/models.ts:15-24` | browse-options doc block |
+| `with-asset-browse.feature.ts:52, 74-77, 148-150, 167-169` | four comment blocks describing folders as a listable, pageable stream |
+
+---
+
+## Tests that must be **inverted**, not just added
+
+Flagged here because it sharpens what "Red" means for the TDD gate — existing tests assert exactly
+the behavior being withdrawn and will fail *by design* once rewritten.
+
+> **Corrected during `/speckit-implement`**: this list said four. The real count is **six**. Two more
+> surfaced only once the source changed, both in `angular-form-bridge.spec.ts`:
+> `should report a folder with its path as the url` (deleted — a folder can no longer be picked, so
+> there is no selection to report) and the `folder` case of
+> `should not attach contentlet-only fields to a %s` (case removed, `link` kept). A spec-file grep
+> for the withdrawn *behavior* would have caught them; a grep for the withdrawn *types* did not.
+
+| Test | File:line | Current assertion | Becomes |
+|---|---|---|---|
+| `should ask for folders when the caller does` | `dot-asset-picker.store.spec.ts:438` | `$request().showFolders` is `true` | `false` — and the test name must change with it |
+| `should no longer pin the folder cursor to zero` | `:475` | `$request().folderCursor` is `5` | removed; the key is no longer sent |
+| `should stop asking for folders once there are no more` | `:501` | `true` on page 1, `false` on page 2 | removed; there is no folder stream to exhaust |
+| `should map folder and link kinds to browse options` | `angular-form-bridge.spec.ts:957` | `{ showFolders: true, showLinks: true }` | `showLinks` only, plus a warning for the unknown `folder` kind |
+
+Two more that must keep passing **unchanged**, as the regression guard:
+
+- `should keep Next reachable while menu links remain` (`:640`)
+- `should settle on the exact total once every stream is exhausted` (`:658`)
diff --git a/specs/37366-assetpicker-list-content-only/spec.md b/specs/37366-assetpicker-list-content-only/spec.md
new file mode 100644
index 000000000000..298a34fb1f22
--- /dev/null
+++ b/specs/37366-assetpicker-list-content-only/spec.md
@@ -0,0 +1,354 @@
+# Issue Resolution Specification: Asset Picker — list panel shows content only, folders navigated through the sidebar tree
+
+**Feature Branch**: `nicobytes/37366-asset-picker-list-panel-should-show-content-only-with-folders-navigated-through-the-sidebar-tree`
+
+**Created**: 2026-09-03
+
+**Status**: Draft — open question settled 2026-09-03 (shape **(a)**, see [Decision](#decision-settled-2026-09-03))
+
+**Type**: Issue / Bug Resolution
+
+**Related GitHub Issue**: [#37366](https://github.com/dotCMS/core/issues/37366) — split from [#37174](https://github.com/dotCMS/core/issues/37174) (finding 9), source [#37207](https://github.com/dotCMS/core/issues/37207) (PR [#37273](https://github.com/dotCMS/core/pull/37273)), epic [#36702](https://github.com/dotCMS/core/issues/36702)
+
+**Input**: User description: "37366-asset-picker-list-panel-should-show-content-only-with-folders-navigated-through-the-sidebar-tree — the field in question is `dotCMS/src/main/webapp/WEB-INF/velocity/static/content/file_browser_field_render_new.vtl`"
+
+---
+
+## ⚠ Premise Corrections
+
+Two statements in the issue body do not survive verification against `main`. Both matter, because
+one of them is the whole argument for the alternative shape.
+
+### 1. `forwardTo` is a **Vanity URL** field, not an HTML-page field
+
+The issue says *"the `forwardTo` field on an HTML page is the shipped case"*. `forwardTo` is a
+field of the **Vanity URL** content type:
+
+- `VanityUrlContentType.java:42` — `FORWARD_TO_FIELD_VAR = "forwardTo"`
+- `VanityUrlContentType.java:80-83` — its custom field is seeded with
+  `$velutil.mergeTemplate('/static/content/file_browser_field_render.vtl')`
+- `file_browser_field_render.vtl:1-5` — that template is only a dispatcher: it parses
+  `file_browser_field_render_new.vtl` when `$structures.isNewEditModeEnabled()` is true, and
+  `file_browser_field_render_old.vtl` otherwise
+
+The HTML page's own custom field is `redirecturl`, rendered by `redirect_custom_field_new.vtl`,
+which asks for `kinds: ["page", "link"]` (`redirect_custom_field_new.vtl:57`) — **no folder**. It is
+therefore unaffected by the `kinds` change and only needs a no-regression check.
+
+**Consequence for the fix**: the Java side changes nothing. `VanityUrlContentType` names the
+*dispatcher*, not `_new.vtl`, so editing `_new.vtl` needs no Java edit and no content-type
+migration.
+
+### 2. Shape (a) does **not** remove a capability the legacy Browser Selector had
+
+The issue offers shape (a) with the caveat that it *"removes a capability the legacy Browser
+Selector had"*. Neither legacy picker ever listed a folder as a selectable row, nor returned one:
+
+| Legacy picker | List content | Returned value |
+|---|---|---|
+| `DotBrowserSelectorComponent` (Angular, still live for Dojo hosts per #37132) | `content.data: DotCMSContentlet[]` (`browser.store.ts:60-64`) — its `folders: TreeNodeItem[]` field is the **sidebar tree**, not list rows | `selectedContent: DotCMSContentlet \| null` (`browser.store.ts:65`) |
+| `dotcms.dijit.FileBrowserDialog` (Dojo, used by `file_browser_field_render_old.vtl:38-39`) | files, via `onFileSelected` | a file; folders live in its `foldersTree` (`FileBrowserDialog.js:164,184-189`) |
+
+A folder became pickable for the first time on **2026-08-31**, when #37273 merged (`81139a83717`),
+and only behind the non-default new-Edit-Content flag. So shape (a) **restores** the behavior both
+legacy pickers had — which is exactly what QA asked for — rather than withdrawing something
+customers have.
+
+There is also no loss of the underlying use case. `forwardTo` keeps its free-text input
+(`file_browser_field_render_new.vtl:38-44`), whose `change` handler writes straight to the field
+value. A Vanity URL can still forward to a folder path; the path is **typed**, not **picked**. What
+goes away is picking one from the browser.
+
+---
+
+## Decision (settled 2026-09-03)
+
+The issue makes one question blocking: *does "pick a folder path" survive as a capability at all?*
+It is settled as **shape (a)**:
+
+> **`'folder'` is dropped from the browse contract.** It leaves `DotBrowserItemKind`, leaves the
+> `DotBrowserSelection` union, and leaves `file_browser_field_render_new.vtl`'s `kinds`. A folder
+> becomes unpickable, and the sidebar tree is the only place folders appear.
+
+And its follow-on: **the picker's folder paging stream is removed**, not kept dormant —
+`folderCursor` / `hasMoreFolders` and their contribution to `$totalRecords` go, so the paging model
+carries no dead stream. The `/api/v1/drive/search` request keeps sending `showFolders: false`,
+because the endpoint takes the flag.
+
+**Shape (b) is rejected, not deferred.** (b) kept the capability and moved it onto an explicit
+"Select this folder" affordance in the tree. It is rejected because Premise Correction 2 removes its
+justification — there is no legacy capability to preserve — while it costs new UI surface, new UX
+design, and a contract that keeps advertising a kind whose selection path is unlike every other
+kind's.
+
+**Who settled it**: the issue owner, during refinement on 2026-09-03. The issue's AC asks for
+UX/Product sign-off before any code; **this spec PR is that record** — approval of PR 1 is the
+sign-off. If UX/Product prefers (b), this spec must be re-approved before `/speckit-plan` runs. See
+[Assumptions](#assumptions).
+
+---
+
+## Problem Statement *(mandatory)*
+
+The Asset Picker's right-hand list panel lists the current folder's **subfolders as selectable
+rows**, each with a single-select radio, mixed in with files and pages — and a folder can be
+returned as the picked value. QA expects the list to carry content only, with folders reached
+through the sidebar tree, which is how every dotCMS file browser before this one behaved.
+
+This is **not a defect in #37207's implementation**. That issue's own acceptance criteria asked for
+this behavior — *"The picker can list and return folders when the caller asks for them"* — and its
+post-merge plan covers folder selection as TC-003, which **passed**. TC-001 failed on QA disagreeing
+with the design. What changes here is the decision, not a regression.
+
+**Severity / Impact**: Medium — some functionality impacted.
+
+- **Who**: back-end users editing a content item whose custom field renders
+  `file_browser_field_render_new.vtl` — in the shipped product, the Vanity URL `forwardTo` field —
+  with the new Edit Content experience enabled (not the default).
+- **How badly**: no data loss and no error. A confusing list, and a field value that can be set to a
+  folder path by picking a row that QA considers navigational.
+- **How often**: every time that picker is opened. Scope is narrow: `DotAssetPickerConfig.browse`
+  keeps folder listing opt-in (`store/models.ts:26-32`), and File, Image, video and audio fields
+  never set it — so only `openBrowserModal` callers are affected, and today exactly one shipped
+  template asks for folders.
+
+## Reproduction *(mandatory)*
+
+**Environment**: latest from `main` (folder listing arrived with `81139a83717`, 2026-08-31); new Edit
+Content experience enabled so `$structures.isNewEditModeEnabled()` is true; any site; any modern
+browser.
+
+**Steps to Reproduce**:
+
+1. Enable the new Edit Content experience.
+2. Edit (or create) a **Vanity URL** content item — its `forwardTo` custom field renders
+   `file_browser_field_render_new.vtl` through the `file_browser_field_render.vtl` dispatcher.
+3. Click **click-here-to-browse** to open the Asset Picker.
+4. Look at the right-hand list panel.
+5. Select a folder row and confirm.
+
+**Expected Behavior**: the list panel carries content only — files and pages, no folder rows.
+Folders are reached by navigating the sidebar tree, and the list reflects the tree's selected folder.
+
+**Actual Behavior**: step 4 — the current path's subfolders are listed as rows, each with a
+selectable radio, interleaved with files and pages. Step 5 — the folder's path is written into the
+field. No error message; the behavior is the shipped template's own request honored end to end.
+
+**Reproducibility**: always, for any `openBrowserModal` caller whose `kinds` includes `'folder'`.
+Never for any other entry point.
+
+> 🎥 Video on the [TC-001 result](https://github.com/dotCMS/core/issues/37207#issuecomment-5486135180).
+
+## Scope of Investigation *(mandatory)*
+
+- **Affected area**: new Edit Content → the custom-field VTL bridge
+  (`DotCustomFieldApi.openBrowserModal`) → the Asset Picker's browse mode and its list panel.
+
+- **Suspected surface**: **modern only**, and almost entirely frontend.
+  - `core-web/libs/edit-content-bridge` — the published browse contract and its translation into
+    the picker's vocabulary.
+  - `core-web/libs/ui/src/lib/components/dot-asset-picker` — the browse options type and the
+    request/paging feature.
+  - One backend-owned *artifact*, but no backend *code*: the Velocity template
+    `dotCMS/src/main/webapp/WEB-INF/velocity/static/content/file_browser_field_render_new.vtl`.
+  - No `com.dotmarketing.*` and no `com.dotcms.*` Java change. No REST resource, no `@Schema`, no
+    `openapi.yaml` regeneration, no DB schema, no ES mapping.
+
+- **Related known decisions**:
+  - **#37132's governing rule stays in force** — *Dojo/old-editor host → the picker already there;
+    Angular host → the new AssetPicker*. The legacy `dot-browser-selector` subtree is untouched by
+    this fix; it is only cited above as evidence of prior behavior.
+  - **#37112's `showLinks` / `linkCursor` paging stays** — only the folder stream is withdrawn.
+  - `/speckit-plan` formally consults `dotCMS/platform-adrs`; no ADR is known to bear on this, and
+    this spec proposes none.
+
+## Root-Cause Hypothesis
+
+**There is no code defect.** The behavior is the shipped template's own request travelling faithfully
+through every layer:
+
+| # | Layer | Location | What it does |
+|---|---|---|---|
+| 1 | Velocity template | `file_browser_field_render_new.vtl:23` | asks for `kinds: ["file", "page", "folder"]` |
+| 2 | Browse-config builder | `angular-form-bridge.ts:665` | `kinds.includes('folder')` → `{ showFolders: true }` |
+| 3 | Picker request | `with-asset-browse.feature.ts:129-130, 169` | forwards `showFolders` on every `/api/v1/drive/search` call |
+| 4 | Endpoint | `/api/v1/drive/search` | returns folders in `response.list` |
+| 5 | Picker store | `with-asset-browse.feature.ts` (`patchState({ items: response.list })`) | folders land in `items` next to contentlets — `items` is typed `DotContentDriveBrowseItem[]` (`store/models.ts:244`) and `selectedAsset` was widened from `DotCMSContentlet` for exactly this (`store/models.ts:297-305`) |
+| 6 | List component | `dot-folder-list-view.component.html:106-109` | renders a `p-tableRadioButton` for **every** row it is handed |
+
+Two consequences follow, and they shape the fix:
+
+- **`DotFolderListViewComponent` has no notion of a navigational, non-selectable row.** `items` is
+  one flat list and the radio is unconditional. So "list content only" cannot be achieved by styling
+  or disabling rows; it has to be enforced **upstream of the list**, in what the picker requests.
+- **Turning it off touches the published contract, not just the picker.** `DotBrowserItemKind`
+  (`asset-browser.interface.ts:19`) advertises `'folder'`, `DotBrowserFolderSelection`
+  (`:132-135`) is a member of the returned union, and the folder paging stream
+  (`with-asset-browse.feature.ts:85, 129-130, 151, 211-212`) exists to serve it.
+
+The fix is therefore to **withdraw the capability from the contract**, per the
+[Decision](#decision-settled-2026-09-03) — not to patch a layer.
+
+## Fix Scope & Non-Goals *(mandatory)*
+
+**In scope**:
+
+1. **The published browse contract** — `libs/edit-content-bridge/src/lib/interfaces/asset-browser.interface.ts`
+   - drop `'folder'` from `DotBrowserItemKind` (`:19`)
+   - remove `DotBrowserFolderSelection` (`:132-135`) and its arm of the `DotBrowserSelection` union
+   - update the `kinds` TSDoc so a template author reads what the picker will actually do
+2. **The bridge** — `libs/edit-content-bridge/src/lib/bridges/angular-form-bridge.ts`
+   - remove the `showFolders` opt-in from `browseOptionsFor` (`:665`)
+   - remove `kindOf`'s `item['type'] === 'folder'` branch (`:678-679`)
+   - remove the `kind === 'folder'` arm of the selection mapper (`:724`)
+   - decide how a stale template that still passes `'folder'` is handled at runtime (see
+     [Regression Risk](#regression-risk-mandatory) — a `console.warn`, mirroring the existing
+     `link` + `mimeTypes` warning at `:653-660`, is the candidate)
+3. **The picker's browse options** — `libs/ui/src/lib/components/dot-asset-picker/store/models.ts`
+   - remove `showFolders` from `DotAssetPickerBrowseOptions` (`:26-32`)
+4. **The picker's paging model** — `.../store/features/with-asset-browse.feature.ts`
+   - drop `hasMoreFolders` from `$totalRecords` (`:85`) — the union becomes content + links
+   - remove the `showFolders` computation (`:129-130`) and send a constant `showFolders: false`
+   - remove `folderCursor` from the request (`:151`) and `folderCursor` / `hasMoreFolders` from the
+     page bookmark (`:211-212`) and from the bookmark's type
+   - update the comments that describe folders as a listable, pageable stream (`:52, 74-77,
+     148-150, 167-169`)
+5. **The shipped template** — `file_browser_field_render_new.vtl:23` → `kinds: ["file", "page"]`
+6. **Tests**, per [Acceptance & Verification](#acceptance--verification-mandatory)
+
+**Explicitly out of scope / non-goals**:
+
+- **`DotFolderListViewComponent` is not touched.** No non-selectable row mode, no conditional radio,
+  no folder-row filtering inside the list. It stays a dumb renderer of whatever it is handed — this
+  is an explicit AC of the issue, not just a preference.
+- **The sidebar tree is not touched.** Folder navigation already works and must keep working
+  unchanged.
+- **Content Drive is not touched.** Its own store legitimately lists folders
+  (`dot-content-drive.store.ts:149-172`); it is a different store with a different product intent.
+  Its `showFolders` logic and its specs stay exactly as they are.
+- **The legacy `dot-browser-selector` subtree is not touched**, and #37132's per-host picker routing
+  is not revisited.
+- **`showLinks` / `linkCursor` (#37112) is not touched.** The link stream stays, and stays paged.
+- **Shape (b) is not built** — no "Select this folder" affordance, in this fix or as follow-up work
+  it implies. It is rejected, and the rejection is recorded above so it is not re-litigated.
+- **`forwardTo`'s free-text input is not touched.** Typing a folder path stays supported; only
+  picking one from the browser goes away.
+- **`redirect_custom_field_new.vtl` is not edited.** It asks for `["page", "link"]` already; it is
+  only re-verified.
+- No Java, no REST contract, no `openapi.yaml`, no DB schema, no ES mapping, no content-type
+  migration.
+
+## Regression Risk *(mandatory)*
+
+- **Blast radius**:
+  - **Paging / the paginator — the real risk.** `$totalRecords` (`with-asset-browse.feature.ts:79-89`)
+    claims one page beyond whenever *any* stream reports more, precisely so a stream that outlives
+    the others stays reachable. Narrowing the union from three streams to two must not break the
+    surviving link stream: #37207's **TC-007** — *"the paginator stays reachable while any stream
+    still has more"* — must be re-verified with the folder stream in its final, removed state.
+  - **`DotAssetPickerBrowseOptions.showFolders` removal is a compile-time break** for any caller
+    that sets it. Verified callers today: the bridge (`angular-form-bridge.ts:665`) and spec
+    fixtures (`asset-picker-config.spec.ts:229`, `dot-asset-picker.store.spec.ts:103` and its
+    `showFolders` suites). No production caller outside the bridge.
+  - **`DotBrowserItemKind` is the VTL-facing contract, and TypeScript does not police VTL.** A
+    template that still passes `kinds: ["folder", …]` hands the bridge a string the type no longer
+    admits. The runtime outcome must be benign and *stated*: folders are simply not listed. Whether
+    to also warn is a plan decision; the issue's last AC ("a customer template cannot ask for a kind
+    the picker silently refuses") argues for warning.
+  - The picker's other entry points (File, Image, video, audio) never set `browse`, so their request
+    is unchanged apart from `showFolders` going from a computed `false` to a constant `false`.
+- **Backward compatibility**:
+  - **The browse API is unshipped.** Per `asset-browser.interface.ts:1-11`, the templates that call
+    it (`*_new.vtl`) only render when new Edit Content is enabled, which is not the default. Folder
+    picking has existed only since 2026-08-31 (`81139a83717`). No shipped customer contract breaks.
+  - **Stored field values are unaffected.** A `forwardTo` value that is a folder path — including one
+    picked in the days since #37273 — keeps its value and keeps working: Vanity URL forwarding is
+    server-side and knows nothing about the picker.
+  - **Not rollback-unsafe.** Frontend plus one Velocity template; none of the DB-schema, ES-mapping
+    or API-contract categories apply.
+- **Data considerations**: none. No migration, no repair of existing data.
+
+## Acceptance & Verification *(mandatory)*
+
+- **AC-001** *(the reproduction)*: with new Edit Content enabled, opening the Asset Picker from a
+  Vanity URL's `forwardTo` field lists **no folder rows** — files and pages only. Steps 4 and 5 of
+  the reproduction no longer produce the actual behavior.
+- **AC-002** *(every caller, including a stale one)*: a caller that passes `kinds: ['folder', …]`
+  gets no folder rows and no crash; the resulting `/api/v1/drive/search` request carries
+  `showFolders: false`. The behavior for such a caller is documented, not silent.
+- **AC-003** *(navigation preserved)*: folders stay fully navigable through the sidebar tree, and
+  the list keeps reflecting the tree's selected folder's contents.
+- **AC-004** *(enforced upstream)*: the change is enforced where the request is built — the browse
+  contract and the picker's browse config. `dot-folder-list-view.component.ts/.html` carry **no
+  diff**; no row is hidden or disabled to achieve this.
+- **AC-005** *(the decision, recorded before code)*: shape **(a)** is implemented — `'folder'` is
+  absent from `DotBrowserItemKind`, from the `DotBrowserSelection` union, and from
+  `file_browser_field_render_new.vtl:23`. Shape (b) is recorded as rejected with its rationale
+  (this document), so the alternative is not re-opened during implementation.
+- **AC-006** *(no dead paging stream)*: the picker's folder paging stream is **removed** —
+  `folderCursor` and `hasMoreFolders` appear nowhere in the picker store, and `$totalRecords` unions
+  content + links only. The request still sends `showFolders: false`, because the endpoint requires
+  the flag.
+- **AC-007** *(both shipped templates verified end to end)*:
+  - `file_browser_field_render_new.vtl` — `kinds` changed; browse, navigate, pick a file, pick a
+    page, save, reopen.
+  - `redirect_custom_field_new.vtl` — unchanged `["page", "link"]`; browse, pick a page, pick a
+    menu link, save, reopen. No folder rows there either (there never were).
+- **AC-008** *(contract docs)*: the `openBrowserModal` TSDoc states the resulting behavior — folders
+  are navigation-only, cannot be requested, and cannot be returned — so a template author cannot ask
+  for a kind the picker refuses.
+
+### Verification method
+
+Frontend-only change, so the layer-appropriate types are **Jest/Spectator unit specs** plus manual
+end-to-end checks of the two templates. Per Constitution Principle V this is stated, not silently
+assumed: **integration, Postman and Karate do not apply** — there is no Java, no REST endpoint and
+no server behavior change; the one backend artifact is a Velocity template whose only assertion is
+the `kinds` literal it passes to the frontend.
+
+**Specs to add / update** (regression coverage the issue asks for):
+
+| Spec | Change |
+|---|---|
+| `libs/ui/.../dot-asset-picker/store/dot-asset-picker.store.spec.ts` | **new**: no folder row reaches `items` for a `kinds: ['folder', …]` caller. **Update**: the `showFolders` suites (`:103`, `:248-275`, `:319`, `:441`, `:476`, `:514-521`) to the final shape |
+| `libs/ui/.../dot-asset-picker/asset-picker-config.spec.ts` | remove the `showFolders: true` fixture (`:229`) |
+| `libs/edit-content-bridge/.../angular-form-bridge.spec.ts` | update the `openBrowserModal` suite (`:945`, `:961`); **add**: `kinds: ['folder', 'file']` produces no `showFolders`, and the documented handling for the now-unknown kind |
+| paging coverage (picker store spec) | **guards #37207 TC-007**: with the folder stream removed, Next stays reachable while only `hasMoreLinks` is true, and the last page reports an exact total |
+
+**Commands**:
+
+```bash
+cd core-web
+pnpm nx test ui --testPathPatterns=dot-asset-picker
+pnpm nx test edit-content-bridge --testPathPatterns=angular-form-bridge
+pnpm nx lint ui && pnpm nx lint edit-content-bridge
+```
+
+**Manual**: new Edit Content enabled → Vanity URL `forwardTo` (AC-001, AC-003, AC-007) and HTML page
+`redirecturl` (AC-007).
+
+**TDD gates** (Constitution V, non-negotiable): the specs above are written first, approved by the
+developer, and confirmed **failing (Red)** before any implementation lands. `/speckit-tasks` orders
+each story tests → approval gate → Red gate → implementation.
+
+## Assumptions
+
+- **The blocking decision is settled with the issue owner, not yet with UX/Product.** Shape (a) was
+  chosen on 2026-09-03 during refinement, on the strength of
+  [Premise Correction 2](#2-shape-a-does-not-remove-a-capability-the-legacy-browser-selector-had).
+  This spec PR **is** the record the issue's AC asks for; its approval is the sign-off. If UX/Product
+  prefers (b), the spec is re-approved before `/speckit-plan` runs — no code is written against an
+  unapproved shape.
+- **No customer VTL outside this repo passes `kinds: ['folder']`.** Assumed because the browse API is
+  unshipped behind a non-default flag and only three days old. If a customer template does, it
+  degrades to "no folder rows" — which is the intended product behavior anyway, so the assumption
+  being wrong changes nothing but the warning's audience.
+- **`/api/v1/drive/search` keeps requiring `showFolders`**, so the picker keeps sending a constant
+  `false` rather than omitting the key. To be confirmed against the endpoint during `/speckit-plan`;
+  if the flag is optional, omitting it is preferred.
+- **Reproduction environment is "latest from `main`"**, as the issue states, with
+  `$structures.isNewEditModeEnabled()` true. Behavior on the default (flag off) path is not in
+  question — `file_browser_field_render_old.vtl` never listed folders.
+- **QA's TC-003 for #37207 ("the picker can return a folder") is superseded by this decision** and is
+  expected to be retired or inverted in that issue's post-merge plan, not treated as a conflicting
+  requirement.