Summary

Destructive slash commands execute immediately with zero confirmation:

By contrast, dangerous terminal commands go through a proper 4-choice approval flow (tools/terminal_tool.py). Slash commands that destroy conversation data have zero safety net.

/rollback is the only destructive command that creates a recovery snapshot first.

Suggested fix

Add a confirmation prompt before destructive commands:

elif cmd_lower == "/clear":
    confirm = input("Clear screen and reset conversation? [y/N] ")
    if confirm.lower() == "y":
        self.console.clear()
        self.conversation_history = []
        self.show_banner()

With options:

• --force flag to skip confirmation (for scripting)
• Config key tui.confirm_destructive: false to disable globally
• Or a simple y/N prompt (less intrusive than the full approval flow)

Impact

• Severity: High — accidental /clear or /reset loses entire conversation
• Risk: Medium — changes muscle-memory for users who type /new<enter> quickly
• Effort: Small

Ref

UX audit item #17.