iterating-with-ai-and-mcp

What problem does it solve?

This Skill enables an AI agent to drive a HotSwan MCP loop that edits Kotlin files, triggers hot reloads, captures device screenshots, and iterates UI toward a mock, reducing manual debugging effort.

Core Features & Use Cases

• Seven MCP tools: exposes hotswan_get_status, hotswan_reload, hotswan_take_screenshot, hotswan_start_snapshot, hotswan_stop_snapshot, hotswan_select_variant, and hotswan_build_and_install to power end-to-end UI iteration.
• Canonical iteration loop: edit → reload → screenshot → evaluate → decide, with optional snapshot rollback for safe experimentation.
• Fallback safety: detects schema changes and uses a full install path when hot reload is not sufficient.
• Use cases: QA automation, AI-assisted UI tuning against design mocks, and rapid visual iteration for Compose screens.

Quick Start

Instruct the agent to connect to the HotSwan MCP server, verify watcher state is WATCHING, start a snapshot, edit the target Kotlin file, trigger a reload, capture a screenshot, and iterate until the result matches the design intent.