Agentcad Skill Release 0.4.0

This Skill addresses the pain point where AI agents cannot directly perform 3D modeling. In practical development, agents or developers need to quickly generate 3D models for prototyping or product design, but traditional CAD software is complex to operate and lacks natural language interaction with AI. AgentCAD combines Python scripting capabilities with AI agents, allowing users to simply describe their requirements, and the AI can automatically generate compliant 3D geometry.

Usage is straightforward: you just need to install agentcad (pip install agentcad), then initialize the project directory via agentcad init. After that, you can have the AI agent write build123d or CadQuery scripts—no need to manually import libraries, as build123d primitives and the show_object function are pre-injected. Execute scripts using the agentcad run script.py --output label command, which supports a dry-run mode to preview metrics like volume and dimensions. A formal run will generate a STEP file, a PNG render, an interactive 3D viewer (viewer.html), and a version comparison diagram. The AI agent can also use agentcad measure to measure feature dimensions, agentcad diff to compare version differences, and agentcad inspect to check for invalid geometry.

It is ideal for teams or individuals who need to collaborate with AI for parametric 3D design. Especially for developers already using build123d or CadQuery as their modeling engine, it integrates seamlessly into AI workflows. Product designers, mechanical engineers, and agent developers can rapidly iterate designs through natural language, saving time on manual scripting. It is also suitable for educational scenarios, where AI assists students with 3D modeling assignments.

It is recommended to clearly inform the AI agent about agentcad's usage in the prompt, especially the script writing rule (must call show_object). Note that it primarily relies on Python 3.10-3.12 environments and can only execute build123d or CadQuery syntax, not other CAD scripts. For complex assemblies, it is advised to first dry-run to confirm geometry validity before a formal run. Additionally, each run creates a new version; manage multiple iterations wisely using the --output label.