This skill turns natural-language descriptions or reference images into build123d Python code, runs that code, and renders the resulting 3D geometry in a local browser viewer. It supports three modes: text-driven generation, image-driven generation (where the agent inspects a provided photo or drawing and researches real-world dimensions), and an edit-apply loop for incremental changes submitted from the viewer. The skill also auto-arms a periodic wakeup to collect and apply viewer edits.
Use this skill when a user asks the agent to create or modify a 3D part: 'render a bracket', 'model from image', 'create a gear', or 'apply pending edits' from the viewer. It is intended for CAD-style geometry (mechanical parts, brackets, standard hardware) rather than organic mesh sculpting.
has_scripts=false) but the skill provides precise instructions to create viewer/models/script.py and run it in the skill venv.Best used by code-capable agents that can run Python and shell commands (Claude/Code-style agents, Codex/Copilot workflows, or any agent with Bash + Python execution).
A 3D model generation skill using build123d with three modes (text, image, edit). The SKILL.md is comprehensive with excellent build123d reference documentation and clear mode detection logic. No bundled scripts to test. The edit-queue pattern for iterative model refinement is well-designed. Minor gaps: no cleanup instructions for the viewer server, limited error state documentation beyond setup failure. Solves a genuinely useful problem with broad appeal to developers and designers.
Well-crafted skill for 3D model generation. The edit-queue pattern (pending/processed directories) is elegant for iterative workflows. No security concerns — localhost-only viewer, no exfiltration, no destructive commands. Broad Bash/WebSearch/WebFetch tool access is justified by the skill's needs. The build123d quick reference embedded in SKILL.md is a nice touch for onboarding.