Confirmation Prompt

Demo

Features

• Pending action queue with approve/deny controls
• Risk level indicators: low (◇), medium (◆), high (▲)
• Tool type badges with semantic coloring
• Status tracking: pending (spinner), approved (✓), denied (✗), auto-approved (⚡)
• Expandable action details
• High-risk warnings for destructive operations
• Batch approval ("approve all") capability

Keybindings

• j / k — Navigate actions
• y — Approve selected action
• n — Deny selected action
• a — Approve all pending
• ⏎ — Toggle action details
• r — Reset demo

Research Context

Inspired by Vercel AI Elements (Confirmation component with accepted/rejected/request states), CopilotKit (CoAgents human-in-the-loop), and assistant-ui (approval workflows). Human-in-the-loop is a critical pattern for agentic AI — risk assessment is a novel addition.