
Forgecad Reconstruction Benchmark
You're in a prepared workspace with a CAD reference asset and a fixed submission path, and your job is to write parametric ForgeCAD source code that reconstructs it. This is the benchmark variant of the reconstruction skill, so you've got sandbox limits and a tight time budget. The rules are strict: no mesh imports, no asset embedding tricks, no external references. You inspect the reference evidence (dimensions, volume, section views), write a brief, then build coarse to fine using the local CLI wrapper. The real discipline here is shipping a runnable parametric approximation instead of burning your budget on perfect investigation. Clean parametric code that approximates the shape beats an unreadable mesh clone every time.
- GitHub stars
- 851
- last updated
- Jun 10, 2026
Install
$ npx -y skills add kostard/forgecad-public-kit --skill forgecad-reconstruction-benchmark --agent claude-codeInstalls into .claude/skills of the current project.
Files
You're in a prepared workspace with a CAD reference asset and a fixed submission path, and your job is to write parametric ForgeCAD source code that reconstructs it. This is the benchmark variant of the reconstruction skill, so you've got sandbox limits and a tight time budget. The rules are strict: no mesh imports, no asset embedding tricks, no external references. You inspect the reference evidence (dimensions, volume, section views), write a brief, then build coarse to fine using the local CLI wrapper. The real discipline here is shipping a runnable parametric approximation instead of burning your budget on perfect investigation. Clean parametric code that approximates the shape beats an unreadable mesh clone every time.
Browse the repository on GitHubSponsors
Details
- Repository
- github.com/kostard/forgecad-public-kit
- Categories
- DevOps & CI/CD, Design & UI/UX
- License
- MIT
- First seen
- Jun 11, 2026