majesticway-task-creation

What problem does it solve?

This skill eliminates the time-consuming and error-prone process of manually formatting AEC benchmark tasks to meet Harbor conventions, ensuring consistency and evaluability for AI agent testing on real construction project documents.

Core Features & Use Cases

It provides end-to-end guidance for designing task instances across all scopes (intrasheet, intradrawing, intraproject) and task families, including defect injection for error detection tasks, navigation task design, verifier script authoring, and local validation. For example, AEC AI researchers can use this skill to quickly generate a new test instance that checks if an agent can detect a mismatched wall cavity dimension between a construction drawing and employer's requirements document.

Quick Start

Use this skill to create a new defect-detection task for the MajesticWay ground floor plan by selecting a source PDF, defining an injected dimension error, and generating the required task.toml, instruction.md, and test.sh files following the provided Harbor format guidelines.