CAD Skills Library
A skills library for CAD, robotics, and hardware design agents to generate, inspect, source, slice, and hand off CAD and robot-description artifacts.
Evidence shows a skills library but no explicit documentation on permission management, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, rollback, or source attribution. No malicious behavior detected, but implementation details are lacking. Scores are low due to insufficient evidence.
Repository includes test scripts and CI configuration, indicating some consistency, but no guarantees on dependency availability or detailed failure messages. Self-consistency scored 2 due to tests and docs; dependency availability and failure messages scored 1 due to limited evidence.
README lists multiple skills and installation methods, indicating broad scenarios, but capability boundaries, trigger precision, and environment fit are not explicitly defined. Scores are 2 for basic coverage but lack depth.
Information architecture is clear, installation notes are detailed, examples and benchmarks are rich, but naming stability, known limitations, versioning/changelog, and maintenance responsibility are not well documented. License is MIT, scored 2.
Output usability is supported by examples and benchmarks, marginal value is high, but cost-benefit is not quantified. Output usability and marginal value scored 2, cost-benefit scored 1.
Claims are supported by benchmarks and docs, but cross-source corroboration and separation of facts and inferences are lacking. Scores are 1 for each.
- Security details such as permission management, user confirmation, and data flow transparency are not provided; use with caution.
- Dependency security is not explicit, posing potential supply chain risks.
- External effects (e.g., Bambu Labs printing) lack rollback mechanisms.
What does this agent do, and when should you use it?
CAD Skills is a library of agent skills for generating, inspecting, sourcing, slicing, and handing off CAD and robot-description artifacts from local project files. It includes multiple focused skills: the CAD skill creates and edits CAD models with STEP as the main output, plus STL, 3MF, and GLB exports; CAD Viewer provides local browser previews; step.parts finds off-the-shelf STEP parts; DXF creates 2D drawings; URDF, SRDF, and SDF skills handle robot descriptions and simulation; SendCutSend checks files before manufacturing; G-code skill slices mesh files; and Bambu Labs manages print jobs. The library is installed via the Skills CLI or as plugins for Codex and Claude Code. It includes a benchmark suite to evaluate CAD generation capabilities.
The library provides multiple agent skills, each performing specific CAD-related operations: the CAD skill reads natural language or image requests to generate STEP files and can export to STL, 3MF, and GLB; CAD Viewer previews CAD, G-code, and robot files in a local browser; step.parts searches for off-the-shelf STEP parts like screws, bearings, and motors; DXF skill generates 2D drawings from Python sources or CAD geometry; URDF skill creates robot structure files with links, joints, limits, inertials, and meshes; SRDF adds MoveIt planning groups and collision rules; SDF creates simulator models and worlds; SendCutSend checks DXF and STEP files before upload; G-code generates validated .gcode using real slicer CLIs; Bambu Labs dry-runs, uploads, and starts print jobs. Installation is via npx skills install earthtojake/text-to-cad, or plugin marketplaces for Codex (codex plugin marketplace add earthtojake/text-to-cad) and Claude Code (claude plugin marketplace add earthtojake/text-to-cad).
- Mechanical engineers can generate editable STEP models from natural language prompts and export STL for 3D printing.
- Robotics developers can add MoveIt planning groups and collision rules to URDF files for motion planning.
- Hardware designers can source standard STEP parts like screws and bearings to accelerate assembly design.
- Manufacturing engineers can validate DXF files for manufacturability before uploading to SendCutSend.
- Makers can start print jobs on Bambu Lab printers from validated G-code files.
- Educators can demonstrate text-to-CAD capabilities with benchmark examples and inspect outputs.
What are this agent's strengths and limitations?
- Supports multiple output formats (STEP, STL, 3MF, GLB, DXF, G-code) covering design to manufacturing.
- Includes specialized robot description skills (URDF, SRDF, SDF) with MoveIt and simulation integration.
- Comes with a benchmark suite to evaluate CAD generation accuracy.
- Multiple installation paths (Skills CLI, Codex plugin, Claude Code plugin) make integration flexible.
- Primarily targets supported agents (Codex and Claude Code); other platforms may need adaptation.
- Requires Node.js and Python 3.11+, adding deployment complexity.
- Not applicable to non-CAD or non-robotics tasks.
- Documentation lacks specific Windows or macOS installation steps, potentially hampering cross-platform use.
- Some skills (e.g., Implicit CAD) are experimental and may be unstable.
How do you install or deploy this agent?
Run npx skills install earthtojake/text-to-cad for standard installation. For Codex (v0.142.0+), run codex plugin marketplace add earthtojake/text-to-cad and codex plugin add cad@text-to-cad. For Claude Code, run claude plugin marketplace add earthtojake/text-to-cad and claude plugin install cad@text-to-cad.
How do you use this agent?
After installation, skills are invoked via agent prompts. For example, describe a CAD model (e.g., "Create a 100x60x20 mm block with four 8 mm holes") and the agent will use the CAD skill to generate a STEP file. For G-code, provide mesh files and the skill slices using CLI slicers.
FAQ
Which agents does this library support?
What are the dependencies?
How do I use the benchmarks?
git lfs pull --include="benchmarks/**" to hydrate locally. The benchmark table shows examples of CAD tasks.