Auto-Approval Mode (--yes)

Overview

The --yes (or -y) flag enables auto-approval mode, which disables all confirmation prompts and allows the AI to execute operations immediately without asking for permission.

This is useful when you trust the AI completely and want maximum automation without interruption.

Quick Start

# Enable auto-approval with long flag
lsbot --yes router --provider deepseek --api-key sk-xxx

# Enable auto-approval with short flag
lsbot -y router --provider deepseek --api-key sk-xxx

# Combine with debug mode
lsbot --yes --debug router --provider deepseek --api-key sk-xxx

What Changes with --yes?

Without --yes (Default Behavior)

The AI may be cautious and ask for confirmation before executing sensitive operations:

Example conversation:

User: Save this content to README_EN.md

AI: ❌ I haven't actually saved the file yet.
    All operations were read-only (analysis and translation).

    Would you like me to save the updated README_EN.md now?
    Please confirm by saying "Yes, save the file" or "Proceed with save"

With --yes (Auto-Approval Mode)

The AI immediately executes the requested operation without asking:

Example conversation:

User: Save this content to README_EN.md

AI: ✅ File saved successfully to README_EN.md (2,847 bytes written)

Behavior Changes

When --yes is enabled, the AI receives these instructions:

1. Execute file writes immediately - No confirmation needed for creating/updating files
2. Run shell commands directly - No permission prompts for command execution
3. Delete/modify files without hesitation - Trust user intent completely
4. Skip safety prompts - User has explicitly disabled all warnings
5. Only reject impossible/dangerous operations - Like rm -rf / or system-breaking commands

Use Cases

✅ When to Use --yes

• Batch operations: Processing multiple files or tasks automatically
• Trusted environment: Running in your own development environment
• Known safe operations: File edits, documentation updates, code generation
• CI/CD pipelines: Automated workflows where manual approval isn't feasible
• Power users: You understand the risks and want maximum efficiency

❌ When NOT to Use --yes

• Production servers: Direct operations on live systems
• Shared environments: Multi-user systems where mistakes affect others
• Learning/testing: First time using a feature or command
• Uncertain operations: When you're not sure what will happen
• Sensitive data: Operations involving credentials, keys, or private information

Safety Considerations

What --yes Does NOT Disable

Even with auto-approval enabled, the AI will still refuse:

1. Destructive system commands: rm -rf /, mkfs, dd if=/dev/zero
2. Dangerous operations: Deleting critical system files or directories
3. Impossible operations: Tasks that would break the system
4. Out-of-scope actions: Operations outside the AI's capabilities

What You Should Still Monitor

• File system changes (use git status to track)
• Shell command outputs (check logs with --debug)
• Network operations (external API calls, downloads)
• Resource usage (CPU, memory, disk space)

Examples

Example 1: Documentation Update

Without --yes:

$ lsbot gateway --provider qwen --api-key sk-xxx

User: Update the README with the new feature description

AI: I've prepared the updated content. Should I write it to README.md?
    Please confirm.

User: Yes, please save it

AI: ✅ File updated successfully

With --yes:

$ lsbot --yes router --provider qwen --api-key sk-xxx

User: Update the README with the new feature description

AI: ✅ README.md updated with new feature description (3 sections added)

Example 2: Batch File Operations

Without --yes:

User: Rename all .txt files to .md in the docs folder

AI: I found 15 .txt files. Should I proceed with renaming them all?

User: Yes

AI: ✅ Renamed 15 files

With --yes:

User: Rename all .txt files to .md in the docs folder

AI: ✅ Renamed 15 files in docs/ (.txt → .md)

Example 3: Code Generation

Without --yes:

User: Create a new API endpoint for user authentication

AI: I've generated the code for the endpoint. Should I create the new file
    at src/api/auth.go?

User: Go ahead

AI: ✅ Created src/api/auth.go

With --yes:

User: Create a new API endpoint for user authentication

AI: ✅ Created src/api/auth.go with authentication endpoint
    - POST /api/auth/login
    - POST /api/auth/logout
    - Middleware for token validation

Command Reference

Global Flag

lsbot [--yes|-y] <command> [options]

• Flag: --yes or -y
• Scope: Global (works with all commands)
• Default: false (auto-approval disabled)
• Type: Boolean (no value needed)

Combining with Other Flags

# Auto-approval + Debug mode
lsbot --yes --debug router --provider deepseek --api-key sk-xxx

# Auto-approval + Verbose logging
lsbot -y --log verbose router --provider deepseek --api-key sk-xxx

# Auto-approval + Custom debug directory
lsbot --yes --debug --debug-dir /tmp/my-debug router [...]

Environment Variables

Currently, auto-approval can only be enabled via command-line flag. There is no environment variable equivalent.

If you need persistent auto-approval, consider:

1. Creating a shell alias:

   alias lingti='lsbot --yes'
   

2. Using a wrapper script:

   #!/bin/bash
   lsbot --yes "$@"
   

Troubleshooting

AI Still Asks for Confirmation

Possible causes:

1. Flag placed incorrectly (must be before subcommand):

   # ❌ Wrong
   lsbot gateway --yes --provider deepseek --api-key sk-xxx
   
   # ✅ Correct
   lsbot --yes router --provider deepseek --api-key sk-xxx
   

2. Using old binary (rebuild after updating):

   go build -o dist/lsbot .
   

How to Verify Auto-Approval is Enabled

Check the logs at startup:

lsbot --yes --log verbose router [...]

In verbose mode, you should see the system prompt includes:

## 🚀 AUTO-APPROVAL MODE ENABLED

Best Practices

1. Version Control: Always use --yes in git-tracked directories

   • Easy to review changes: git diff
   • Easy to undo mistakes: git reset --hard

2. Start with --debug: First time using auto-approval?

   lsbot --yes --debug router [...]
   

   This shows all operations being performed.

3. Test in Safe Directory: Try auto-approval in a test folder first

   mkdir /tmp/test-auto-approve
   cd /tmp/test-auto-approve
   lsbot --yes router [...]
   

4. Review Regularly: Periodically check what was changed

   git log --oneline --name-only
   git diff HEAD~5
   

5. Use with Specific Tasks: Enable only for specific operations

   # For file operations
   lsbot --yes router --provider deepseek --api-key sk-xxx
   
   # Regular mode for exploratory tasks
   lsbot gateway --provider deepseek --api-key sk-xxx
   

Security Notes

• Not a Security Feature: --yes is about convenience, not security
• Trust Required: Only use when you trust the AI's judgment
• Audit Trail: Always keep logs (--log verbose) when using auto-approval
• Backup First: Have backups before enabling auto-approval on important files
• Gradual Adoption: Start with read-only tasks, then gradually enable for writes

Comparison Table

Related Flags

• --debug: Enable debug mode (shows all operations)
• --log <level>: Set logging level (silent, info, verbose, very-verbose)
• --debug-dir <path>: Set debug screenshot directory

FAQ

Q: Can I set auto-approval as default? A: Currently no. You must explicitly pass --yes each time. This is by design for safety.

Q: Does --yes bypass all security checks? A: No. It only disables confirmation prompts. Dangerous operations (like rm -rf /) are still blocked.

Q: Can I use --yes with the MCP server mode? A: The --yes flag currently only affects the router command. Other commands may not support it.

Q: What if I want to cancel after using --yes? A: Use Ctrl+C to stop the program at any time. This works with or without --yes.

Q: Does --yes make operations faster? A: No. It only removes confirmation prompts. The actual operations take the same time.

See Also

• Browser Debug Mode - Debug browser automation
• Global Flags - All available global flags
• Security Configuration - Configure security settings