Add operation-specific confirmation settings for AI assistant tools

[Zen0space]

Current Behavior
When "Allow running editing tool without asking for confirmation" is enabled, ALL tool operations (read, write, create, delete, etc.) run automatically without user confirmation.

Requested Feature

Implement a whitelist/blacklist system or operation-specific confirmation settings that allow users to customize which operations require confirmation.

Use Case

When working with AI assistants like Claude, users want a smooth workflow where the AI can freely read, edit, and create files without constant interruptions. However, certain operations (delete files, running servers, etc.) should always require explicit user confirmation for safety.

Proposed Solutions

Option 1: Simple Whitelist/Blacklist Input Field

Settings > AI Assistant Tools:
☑️ Allow editing tools without asking for confirmation

Always ask for confirmation when:
┌─────────────────────────────────────────┐
│ delete_path                             │
│ npm start                               │
│ npm run dev                             │
│ sudo                                    │
│ rm -rf                                  │
│ git push                                │
└─────────────────────────────────────────┘
[Add] [Remove]

□ Use tool names (delete_path, terminal)
□ Use command patterns (npm start, sudo *)

Option 2: Operation Categories

Settings > AI Assistant Tools:
☑️ Allow without confirmation:
  ☑️ Read files
  ☑️ Edit files  
  ☑️ Create files
  ☑️ Run terminal commands
  ☐ Delete files/directories

☑️ Always ask for confirmation:
  ☑️ Delete operations
  ☐ Terminal commands with sudo

Option 3: Combined Approach

☑️ Allow editing tools without asking
☑️ Always confirm destructive operations (delete, move)

Custom confirmation rules:
┌─────────────────────────────────────────┐
│ npm run dev                             │
│ npm start                               │
│ npx expo start                          │
│ python -m http.server                   │
└─────────────────────────────────────────┘

Benefits

• Flexibility: Users can customize exactly what they want to block
• Safety: Prevents accidental file deletion and unwanted server starts
• Productivity: Maintains smooth workflow for safe operations
• User control: Granular control with simple text input

Example Use Cases

• Block delete_path to prevent accidental file deletion
• Block npm start, npm run dev to prevent auto-starting development servers
• Block sudo commands for security
• Block git push to prevent accidental deployments
• Allow pattern matching like rm *, sudo *

Implementation Notes

• Support both exact matches and wildcard patterns
• Could work with tool names (delete_path) and command content (npm start)
• Simple text input field with one rule per line
• Default suggestions for common dangerous operations

---

Additional Context:
Many developers use AI assistants for rapid prototyping and would benefit from this safety-focused, customizable approach to tool confirmations.

[etodanik]

This is similar to this:
#32486

An dyes, a very much needed feature. I don't wanna have to click to allow every single cargo build but I also don't want to blanket allow everything.