AI analysis grounded in the code graph — computed facts, not vibes · 2026-10-05T03:03:24Z
Text-to-CAD (branded "CAD Skills") is a library of agent skills — installable modules like skills/cad, skills/cad-viewer, skills/dxf, skills/urdf, skills/sdf, skills/srdf, and skills/step-parts — that let LLM-based coding agents generate, inspect, and export CAD and robot-description artefacts (STEP, STL, 3MF, GLB, URDF, SDF, SRDF) from natural-language or image requests operating on local project files. It targets developers building agentic workflows for CAD, fabrication, and robotics who want their agent to produce and preview real engineering artefacts rather than just code. A companion browser-based viewer (skills/cad-viewer/scripts/viewer/backend/server.mjs) renders local previews of the generated geometry, G-code, and robot files.
The repo went from creation on 2026-04-22 to 9,295 stars, with 291 gained in a single day, and the commit history shows five releases (0.3.5 → 0.3.9) in under a month, indicating active, frequent shipping rather than a one-off launch. The README's badge row (STEP/STL/3MF/URDF/SDF/SRDF export, Python 3.11+) signals broad, concrete tooling claims that plausibly drive adoption among CAD/robotics-adjacent agent builders, and a Discord community link was added in the same window, suggesting deliberate community-building alongside releases. The provided facts don't include external mentions or traffic data, so the exact driver of the single-day 291-star spike can't be confirmed from this evidence alone.
What changed recently, how it's actually built (from the code graph), and whether you should care. Free account — no card, no spam.