Agentcad Skill Release 0.4.0

This Skill addresses the inability of AI agents to directly design, model, and generate three-dimensional objects. In practical development, when users request to "design a 3D part" or "export an STL file," standard AI models lack the capability to invoke CAD tools. Agentcad enables AI to directly write and execute build123d or CadQuery scripts, outputting formats such as STEP, STL, and GLB, along with rendered images and geometric parameters.

Usage is straightforward: simply describe your requirements in natural language, and the AI will automatically generate Python scripts and execute them via the agentcad run command. Start with --dry-run to preview volume and dimensions; once confirmed, proceed with the full run. Each output generates a four-view preview image (preview.png), a diff image highlighting changes from the previous version, and an interactive 3D HTML viewer. You can also use agentcad inspect to check for invalid geometry, agentcad measure to measure features like hole diameters and edge lengths, and even write specifications into spec.json for automatic design validation with agentcad check-spec.

It is well-suited for scenarios requiring rapid prototyping, such as agent development, product design, and educational demonstrations. Teams already using build123d or CadQuery for parametric modeling can leverage AI's conversational capabilities to significantly shorten the iteration cycle from concept to final model. University students can also use it for 3D modeling in graduation projects without manually writing complex scripts.

It is recommended to run agentcad init to initialize the project before first use and review agentcad --help to understand all commands. Note that it primarily relies on Python 3.10-3.12 and requires prior installation of the agentcad package (pip install agentcad). Output files include STEP, PNG, STL, GLB, OBJ, etc., suitable for 3D printing, CAE analysis, or visualization.