feat(xiaohongshu): use CDP DOM.setFileInputFiles for image upload

extension/src/background.ts

@@ -252,6 +252,8 @@ async function handleCommand(cmd: Command): Promise<Result> {
         return await handleCloseWindow(cmd, workspace);
       case 'sessions':
         return await handleSessions(cmd);
+      case 'set-file-input':
+        return await handleSetFileInput(cmd, workspace);
       default:
         return { id: cmd.id, ok: false, error: `Unknown action: ${cmd.action}` };
     }
@@ -579,6 +581,19 @@ async function handleCloseWindow(cmd: Command, workspace: string): Promise<Resul
   return { id: cmd.id, ok: true, data: { closed: true } };
 }
 
+async function handleSetFileInput(cmd: Command, workspace: string): Promise<Result> {
+  if (!cmd.files || !Array.isArray(cmd.files) || cmd.files.length === 0) {
+    return { id: cmd.id, ok: false, error: 'Missing or empty files array' };
+  }
+  const tabId = await resolveTabId(cmd.tabId, workspace);
+  try {
+    await executor.setFileInputFiles(tabId, cmd.files, cmd.selector);
+    return { id: cmd.id, ok: true, data: { count: cmd.files.length } };
+  } catch (err) {
+    return { id: cmd.id, ok: false, error: err instanceof Error ? err.message : String(err) };
+  }
+}
+
 async function handleSessions(cmd: Command): Promise<Result> {
   const now = Date.now();
   const data = await Promise.all([...automationSessions.entries()].map(async ([workspace, session]) => ({

extension/src/cdp.ts

@@ -147,6 +147,48 @@ export async function screenshot(
   }
 }
 
+/**
+ * Set local file paths on a file input element via CDP DOM.setFileInputFiles.
+ * This bypasses the need to send large base64 payloads through the message channel —
+ * Chrome reads the files directly from the local filesystem.
+ *
+ * @param tabId - Target tab ID
+ * @param files - Array of absolute local file paths
+ * @param selector - CSS selector to find the file input (optional, defaults to first file input)
+ */
+export async function setFileInputFiles(
+  tabId: number,
+  files: string[],
+  selector?: string,
+): Promise<void> {
+  await ensureAttached(tabId);
+
+  // Enable DOM domain (required for DOM.querySelector and DOM.setFileInputFiles)
+  await chrome.debugger.sendCommand({ tabId }, 'DOM.enable');
+
+  // Get the document root
+  const doc = await chrome.debugger.sendCommand({ tabId }, 'DOM.getDocument') as {
+    root: { nodeId: number };
+  };
+
+  // Find the file input element
+  const query = selector || 'input[type="file"]';
+  const result = await chrome.debugger.sendCommand({ tabId }, 'DOM.querySelector', {
+    nodeId: doc.root.nodeId,
+    selector: query,
+  }) as { nodeId: number };
+
+  if (!result.nodeId) {
+    throw new Error(`No element found matching selector: ${query}`);
+  }
+
+  // Set files directly via CDP — Chrome reads from local filesystem
+  await chrome.debugger.sendCommand({ tabId }, 'DOM.setFileInputFiles', {
+    files,
+    nodeId: result.nodeId,
+  });
+}
+
 export async function detach(tabId: number): Promise<void> {
   if (!attached.has(tabId)) return;
   attached.delete(tabId);

extension/src/protocol.ts

@@ -5,7 +5,7 @@
  * Everything else is just JS code sent via 'exec'.
  */
 
-export type Action = 'exec' | 'navigate' | 'tabs' | 'cookies' | 'screenshot' | 'close-window' | 'sessions';
+export type Action = 'exec' | 'navigate' | 'tabs' | 'cookies' | 'screenshot' | 'close-window' | 'sessions' | 'set-file-input';
 
 export interface Command {
   /** Unique request ID */
@@ -32,6 +32,10 @@ export interface Command {
   quality?: number;
   /** Whether to capture full page (not just viewport) */
   fullPage?: boolean;
+  /** Local file paths for set-file-input action */
+  files?: string[];
+  /** CSS selector for file input element (set-file-input action) */
+  selector?: string;
 }
 
 export interface Result {

src/browser/daemon-client.ts

@@ -19,7 +19,7 @@ function generateId(): string {
 
 export interface DaemonCommand {
   id: string;
-  action: 'exec' | 'navigate' | 'tabs' | 'cookies' | 'screenshot' | 'close-window' | 'sessions';
+  action: 'exec' | 'navigate' | 'tabs' | 'cookies' | 'screenshot' | 'close-window' | 'sessions' | 'set-file-input';
   tabId?: number;
   code?: string;
   workspace?: string;
@@ -30,6 +30,10 @@ export interface DaemonCommand {
   format?: 'png' | 'jpeg';
   quality?: number;
   fullPage?: boolean;
+  /** Local file paths for set-file-input action */
+  files?: string[];
+  /** CSS selector for file input element (set-file-input action) */
+  selector?: string;
 }
 
 export interface DaemonResult {

src/browser/page.ts

@@ -339,6 +339,22 @@ export class Page implements IPage {
     return Array.isArray(result) ? result : [];
   }
 
+  /**
+   * Set local file paths on a file input element via CDP DOM.setFileInputFiles.
+   * Chrome reads the files directly from the local filesystem, avoiding the
+   * payload size limits of base64-in-evaluate.
+   */
+  async setFileInput(files: string[], selector?: string): Promise<void> {
+    const result = await sendCommand('set-file-input', {
+      files,
+      selector,
+      ...this._cmdOpts(),
+    }) as { count?: number };
+    if (!result?.count) {
+      throw new Error('setFileInput returned no count — command may not be supported by the extension');
+    }
+  }
+
   async waitForCapture(timeout: number = 10): Promise<void> {
     const maxMs = timeout * 1000;
     await sendCommand('exec', {

src/clis/xiaohongshu/publish.test.ts

@@ -8,7 +8,7 @@ import { getRegistry } from '../../registry.js';
 import type { IPage } from '../../types.js';
 import './publish.js';
 
-function createPageMock(evaluateResults: any[]): IPage {
+function createPageMock(evaluateResults: any[], overrides: Partial<IPage> = {}): IPage {
   const evaluate = vi.fn();
   for (const result of evaluateResults) {
     evaluate.mockResolvedValueOnce(result);
@@ -37,10 +37,88 @@ function createPageMock(evaluateResults: any[]): IPage {
     getCookies: vi.fn().mockResolvedValue([]),
     screenshot: vi.fn().mockResolvedValue(''),
     waitForCapture: vi.fn().mockResolvedValue(undefined),
+    ...overrides,
   };
 }
 
 describe('xiaohongshu publish', () => {
+  it('prefers CDP setFileInput upload when the page supports it', async () => {
+    const cmd = getRegistry().get('xiaohongshu/publish');
+    expect(cmd?.func).toBeTypeOf('function');
+
+    const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), 'opencli-xhs-publish-'));
+    const imagePath = path.join(tempDir, 'demo.jpg');
+    fs.writeFileSync(imagePath, Buffer.from([0xff, 0xd8, 0xff, 0xd9]));
+
+    const setFileInput = vi.fn().mockResolvedValue(undefined);
+    const page = createPageMock([
+      'https://creator.xiaohongshu.com/publish/publish?from=menu_left',
+      { ok: true, target: '上传图文', text: '上传图文' },
+      { state: 'editor_ready', hasTitleInput: true, hasImageInput: true, hasVideoSurface: false },
+      'input[type="file"][accept*="image"],input[type="file"][accept*=".jpg"],input[type="file"][accept*=".jpeg"],input[type="file"][accept*=".png"],input[type="file"][accept*=".gif"],input[type="file"][accept*=".webp"]',
+      false,
+      true,
+      { ok: true, sel: 'input[maxlength="20"]' },
+      { ok: true, sel: '[contenteditable="true"][class*="content"]' },
+      true,
+      'https://creator.xiaohongshu.com/publish/success',
+      '发布成功',
+    ], {
+      setFileInput,
+    });
+
+    const result = await cmd!.func!(page, {
+      title: 'CDP上传优先',
+      content: '优先走 setFileInput 主路径',
+      images: imagePath,
+      topics: '',
+      draft: false,
+    });
+
+    expect(setFileInput).toHaveBeenCalledWith(
+      [imagePath],
+      expect.stringContaining('input[type="file"][accept*="image"]'),
+    );
+    const evaluateCalls = (page.evaluate as any).mock.calls.map((args: any[]) => String(args[0]));
+    expect(evaluateCalls.some((code: string) => code.includes('atob(img.base64)'))).toBe(false);
+    expect(result).toEqual([
+      {
+        status: '✅ 发布成功',
+        detail: '"CDP上传优先" · 1张图片 · 发布成功',
+      },
+    ]);
+  });
+
+  it('fails fast when only a generic file input exists on the page', async () => {
+    const cmd = getRegistry().get('xiaohongshu/publish');
+    expect(cmd?.func).toBeTypeOf('function');
+
+    const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), 'opencli-xhs-publish-'));
+    const imagePath = path.join(tempDir, 'demo.jpg');
+    fs.writeFileSync(imagePath, Buffer.from([0xff, 0xd8, 0xff, 0xd9]));
+
+    const setFileInput = vi.fn().mockResolvedValue(undefined);
+    const page = createPageMock([
+      'https://creator.xiaohongshu.com/publish/publish?from=menu_left',
+      { ok: true, target: '上传图文', text: '上传图文' },
+      { state: 'editor_ready', hasTitleInput: true, hasImageInput: true, hasVideoSurface: false },
+      null,
+    ], {
+      setFileInput,
+    });
+
+    await expect(cmd!.func!(page, {
+      title: '不要走泛化上传',
+      content: 'generic file input 应该直接报错',
+      images: imagePath,
+      topics: '',
+      draft: false,
+    })).rejects.toThrow('Image injection failed: No file input found on page');
+
+    expect(setFileInput).not.toHaveBeenCalled();
+    expect(page.screenshot).toHaveBeenCalledWith({ path: '/tmp/xhs_publish_upload_debug.png' });
+  });
+
   it('selects the image-text tab and publishes successfully', async () => {
     const cmd = getRegistry().get('xiaohongshu/publish');
     expect(cmd?.func).toBeTypeOf('function');