mock_websites

What problem does it solve?

This Skill provides a framework to set up and manipulate mock web application state for CUA-Gym tasks, enabling the creation of initial and golden environments and the verification of reward functions.

Core Features & Use Cases

• State Injection: Inject initial and current states into mock web applications via HTTP APIs.
• Task Verification: Write reward scripts to verify task completion against mock web application state.
• Use Case: For tasks like "write a professional greeting in #general" on a mock Slack app, the Skill allows you to programmatically verify the greeting's quality using a combination of direct checks and LLM-based judgment.

Quick Start

Use the mock_websites skill to create an initial state for a mock Slack application by setting up a new channel with the name "general".