What problem does it solve? Testing a Decentraland SDK7 scene normally requires manual launching, walking, and clicking inside the Explorer client. This Skill lets an AI agent drive a running Explorer build through its embedded MCP automation server, so scenes can be edited, hot-reloaded, screenshotted, and verified in-world autonomously. ## Core Features & Use Cases - In-world scene iteration: Edit TypeScript scene code, watch hot-reload, then verify with screenshots, scene-state reads, and log polling until every requirement has visual proof. - Synthetic input and camera control: Teleport, walk, look, click entities, send chat, trigger emotes, and position free or retail camera modes to test interactions and framing. - Performance and content-budget debugging: Measure frame rates, scene content stats, and per-model breakdowns to find what exceeds per-parcel limits. - Use Case: While building a multiplayer maze scene, ask the agent to walk through each trigger area, click every switch, screenshot the results, and confirm the ECS state matches the intended gameplay. ## Quick Start Ask the agent to launch the scene with the MCP flag and verify in the running Explorer that the door opens when the switch is clicked.
