vision-decompose

What problem does it solve?

This Skill solves the frequent issue of LLM CAD agents misinterpreting engineering reference images (drawings, renders, photos, sketches), which leads to wasted time building incorrect parts and running multiple regeneration iterations to fix avoidable errors caused by misread features, missed callouts, or confused additive/subtractive geometry.

Core Features & Use Cases

• Structured Feature Tree Generation: Produces a standardized, reviewable feature tree listing every feature's type, size, position, role, and relationships, so users can catch misreads before any modeling operations begin.
• Guided Image Inspection Workflow: Walks the agent through systematic cropping of reference images at native resolution to read small dimension callouts, identify hidden features on isometric views, and distinguish between additive (boss) and subtractive (pocket/hole) features.
• Use Case Example: When a user shares an engineering drawing of a motor mounting bracket and asks to model it in Onshape, this skill first outputs a complete feature breakdown including all bolt hole sizes, positions, fillet specifications, and derived outline details for user sign-off before any CAD mutations are run.

Quick Start

Use the vision-decompose skill to analyze the attached engineering drawing of the bracket and produce a structured feature tree for your review before starting the build.