physical-verification

What problem does it solve?

This Skill prevents AI from "hallucinating" or falsely claiming UI features work based solely on code changes. It ensures that every UI modification is physically verified with visual evidence, guaranteeing accuracy and preventing "AI slop."

Core Features & Use Cases

• Screenshot Capture: Take precise before-and-after screenshots of the user interface.
• Element Identification: Locate and verify specific UI elements and their states using computer vision.
• Interaction Simulation: Simulate user interactions (e.g., clicks, typing) and visually confirm the expected outcomes.
• Use Case: After implementing a new "Cost Dashboard" button, use this skill to confirm it's visible, in the correct location, clickable, and that clicking it toggles the dashboard's visibility, all backed by visual proof.

Quick Start

1. Take a screenshot of the initial UI state: python -c "from services.chatkit_backend.app.tools.computer_use import get_computer_tool; import json; result = get_computer_tool().screenshot(); print(json.dumps(result))"
2. Verify a UI element's state: from services.chatkit_backend.app.vision import AuditorAgent import asyncio import os auditor = AuditorAgent(api_key=os.getenv("ANTHROPIC_API_KEY")) result = await auditor.verify_ui_element(element_description="CostDashboard button", expected_state="visible, pressable")