Purpose
Automate the process of creating new documents in Feishu (Lark) workspace by using browser automation to navigate the UI, authenticate, and create documents with specified titles and content.
Prerequisites
- Access to Feishu workspace (https://leiniao-ibg.feishu.cn)
- Valid authentication credentials for the workspace
- Browser automation capability via
agent-browser
Workflow
Step 1: Load Browser Automation Skill
Load the office:agent-browser skill using the Skill tool to access browser automation commands.
Step 2: Navigate to Feishu Drive
Open the Feishu Drive homepage:
agent-browser open https://leiniao-ibg.feishu.cn/drive/home/
Wait for page to load:
agent-browser wait --load networkidle
Step 3: Verify Authentication
Take a snapshot to check if already logged in:
agent-browser snapshot -i
If login is required, wait for user to complete authentication manually or handle authentication flow based on the page state.
Step 4: Create New Document
Click the "新建" (New) button (use snapshot to locate the element ref):
agent-browser snapshot -i
# Locate "新建" button ref (e.g., @e1)
agent-browser click @e1
Wait for dropdown menu to appear:
agent-browser wait 1000
Take another snapshot to locate the "文档" (Doc) option:
agent-browser snapshot -i
# Locate "文档" button ref (e.g., @e2)
agent-browser click @e2
Step 5: Select New Document Type
Click the "新建空白文档" (New Doc) option from the submenu:
agent-browser wait 1000
agent-browser snapshot -i
# Locate "新建空白文档" button ref (e.g., @e3)
agent-browser click @e3
Step 6: Wait for New Tab
Wait for the new document to open in a new tab:
agent-browser wait --load networkidle
Check tabs to ensure new document page opened:
agent-browser tab
If multiple tabs exist, switch to the newest tab (usually the last one):
agent-browser tab 2 # Adjust index based on tab list
Step 7: Enter Document Title
The page should automatically focus on the title input field. If the title field is focused by default, type the title directly:
agent-browser type @e1 "Document Title Here"
If not automatically focused, take a snapshot to locate the title input:
agent-browser snapshot -i
# Locate title input ref (e.g., @e1)
agent-browser fill @e1 "Document Title Here"
Step 8: Enter Document Content
Press Tab or click to move to the content area:
agent-browser press Tab
Or locate and click the content editor:
agent-browser snapshot -i
# Locate content editor ref (e.g., @e2)
agent-browser click @e2
Type the document content:
agent-browser type @e2 "Document content goes here..."
For multi-line content, use newlines in the input:
agent-browser type @e2 "First paragraph
Second paragraph
Third paragraph"
Step 9: Verify and Save
Take a final screenshot to verify the document was created successfully:
agent-browser screenshot
Feishu documents auto-save, so no explicit save action is required. The document is now ready to use.
Step 10: Close Browser (Optional)
Close the browser session when done:
agent-browser close
Error Handling
Authentication Issues
If authentication fails or login is required:
- Pause the workflow
- Inform the user that manual login is needed
- Wait for confirmation before proceeding
- Resume workflow after authentication
Element Not Found
If snapshot cannot locate expected UI elements (button refs):
- Take a full snapshot without
-iflag for debugging - Check if UI has changed or language settings differ
- Use semantic locators as fallback:
agent-browser find text "新建" click # Find "新建" (New) button agent-browser find text "文档" click # Find "文档" (Doc) button agent-browser find text "新建空白文档" click # Find "新建空白文档" (New Doc) button
Timeout Issues
If page loading takes too long:
- Increase wait timeout:
agent-browser wait --load networkidle --timeout 10000 - Check network connectivity
- Verify Feishu service availability
Customization
Different Workspace
To use with a different Feishu workspace, replace the URL in Step 2:
agent-browser open https://your-workspace.feishu.cn/drive/home/
Document Templates
To use a specific document template instead of blank document:
- Navigate to template gallery after clicking "Doc"
- Locate and click the desired template
- Proceed with title and content entry
Best Practices
- Session Reuse: For multiple document creations, keep the browser session open and reuse authentication state
- Error Screenshots: Take screenshots at each critical step for debugging
- Wait for UI: Always wait for network idle after navigation to ensure UI elements are loaded
- Explicit Waits: Use explicit waits (e.g.,
agent-browser wait 1000) after clicking dropdown menus
Additional Resources
Browser Automation Reference
For detailed browser automation commands and patterns:
- Load
office:agent-browserskill for complete command reference - See snapshot and interaction patterns in agent-browser documentation
Example Usage
# Complete workflow example
agent-browser open https://leiniao-ibg.feishu.cn/drive/home/
agent-browser wait --load networkidle
agent-browser snapshot -i
agent-browser click @e1 # 新建 button
agent-browser wait 1000
agent-browser snapshot -i
agent-browser click @e2 # 文档 button
agent-browser wait 1000
agent-browser snapshot -i
agent-browser click @e3 # 新建空白文档 button
agent-browser wait --load networkidle
agent-browser tab
agent-browser tab 2 # Switch to new tab
agent-browser type @e1 "My Document Title"
agent-browser press Tab
agent-browser type @e2 "My document content..."
agent-browser screenshot