docs(skills): roadmap — deepened-notch residual re-identified as the terminal stranded-rim case

[andymai]

Iteration-19 decode: the pave-bypass theory is dead (the cone face's section gate is open and it receives all three notch sections). The remaining posBad=6 is the sections arriving with different terminal rims per side (old z=−1.19 rim on the cone, new z=−1.2 rim on the wall planes) — i.e. the already-documented TERMINAL deepened-notch stranded-old-floor-edge case. Updates the row so future sessions start at the detection design (polygon_union) instead of re-deriving this.

---

Summary by cubic

Update skills roadmap to reclassify the residual as the terminal deepened-notch stranded-rim case and drop the pave-bypass theory. Documents the rim mismatch (cone trimmed to old z=−1.19; walls to new z=−1.2), confirms 2.126.12 tool verification, and sets the next step to detect deepened-notches via polygon_boolean::polygon_union without near-coincident false positives.

^{Written for commit c75f232. Summary will update on new commits.}

[greptile-apps]

Greptile Summary

This PR updates a single roadmap tracking entry in .claude/skills/roadmap/SKILL.md with no code changes. It corrects the working theory for the remaining posBad=6 case: the pave-bypass hypothesis is retired (the cone face's section gate is open and receives all three notch sections), and the residual is now definitively identified as the already-documented terminal deepened-notch stranded-old-floor-edge case.

• The entry is updated with tool-verified regression counts (dovetail 9/9, dovetailKey 2/2, fit-offset 2/2, snapClip nozzle nm 13→12) and a raw repro reduction note (37 → 6).
• The fix direction is now pinned to polygon_boolean::polygon_union for merging the old opening's outline into the new one without false-positives on near-coincident faces, pointing future sessions directly to the detection design step.

Confidence Score: 5/5

Documentation-only change to a roadmap skill file with no impact on compiled code, tests, or public interfaces.

The change rewrites a single tracking paragraph in a markdown skill document. No Rust source, Cargo manifests, WASM bindings, or test fixtures are touched, so there is no risk of introducing regressions or breaking anything in the build.

No files require special attention.

Important Files Changed

Filename	Overview
.claude/skills/roadmap/SKILL.md	Documentation-only update: re-identifies the remaining posBad=6 root cause from the pave-bypass theory to the terminal stranded-rim case, adds tool-verified regression counts, and names polygon_union as the required fix primitive.

_{Reviews (1): Last reviewed commit: "docs(skills): roadmap — deepened-notch r..." | Re-trigger Greptile}

[Copilot]

Pull request overview

Updates the .claude skills roadmap entry for the snapClip deepened-notch investigation to reclassify the remaining posBad=6 residual as the already-defined “TERMINAL deepened-notch stranded-old-floor-edge” case (rim mismatch), and removes the now-invalid pave-bypass hypothesis. This helps future debugging sessions start from the correct next-step design work (deepened-notch detection via polygon union) rather than re-deriving prior conclusions.

Changes:

• Rewrites the residual root-cause note to confirm the cone face does receive all three sections and that the issue is a terminal rim mismatch (old vs new rim).
• Adds tool-verification notes and reframes the follow-up work as deepened-notch detection using polygon_boolean::polygon_union.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[github-actions]

WASM Binary Size

File	main	PR	Delta
brepkit_wasm.wasm	5.83 MB	5.83 MB	0 KB (0%)

✅ No size change.