From the rose-pro skin work (rose pro/skin branch). The two worst hour-eaters were silent failures:
- A
[skin] group spanning shoulder-can to shoulder-can blew the 40M grid cap and the root evaluated EMPTY. vcad-render printed only no solids produced — no root name, no reason. vcad info still listed the root as a scene entry, deepening the confusion.
- A group of ONLY thin tilted 3 mm plates (~1.5 cells at r/6) vanished in the SDF and skinned to nothing, again with no message at eval time.
Ask: every refusal/degradation path (grid cap, thin-feature drop, coarsening, --gpu caps) should emit one line naming the root, the cause, and the parameter that fixes it. SkinReport.coarsened exists but nothing surfaces it in the CLI.
🤖 Generated with Claude Code
From the rose-pro skin work (rose pro/skin branch). The two worst hour-eaters were silent failures:
[skin]group spanning shoulder-can to shoulder-can blew the 40M grid cap and the root evaluated EMPTY.vcad-renderprinted onlyno solids produced— no root name, no reason.vcad infostill listed the root as a scene entry, deepening the confusion.Ask: every refusal/degradation path (grid cap, thin-feature drop, coarsening, --gpu caps) should emit one line naming the root, the cause, and the parameter that fixes it.
SkinReport.coarsenedexists but nothing surfaces it in the CLI.🤖 Generated with Claude Code