[tiensonqin]

Summary

Expose row/column/box ~shrink:0. so direct layout children can retain their natural extent without abusing grow:0 (which also changes auto basis in MyGo).

• Add one typed ShrinkValue / shrink wire property: finite nonnegative floats, only row/column/box; omitted/removed means the existing factor 1. Grow and basis are unchanged.
• Map Web to flex-shrink and MyGo to the already-pinned SDK's Element.Shrink. Generic/SwiftUI/Kotlin opt-in is explicitly rejected, including direct Apple/Android standard wire patches; no raw extension escape is added.
• Regenerate canonical schema/MyGo artifacts and capability evidence, and document downstream migration. No SDK, dependency, lockfile, Dune or Logseq consumer changes.

Originally based on public 2475de1e9fecf3c987b3d07617699dd90c21ec73; merge commit 790ca9e0e6a807368a7b61db49021ed11d178f58 incorporates public main 347de2f7a22968e2a54c4504cfa07fe3800f97a6 without rewriting either history. The resolved constructors/schema retain both shrink and focus-within observation, including both unsupported-host guards. Generated artifacts and the three affected capability hashes were regenerated from the combined contract.

Current-head local validation: OCaml @all @web, 235 core tests and 8 portable-service tests; schema tooling 12 tests; Web build + 160 tests; MyGo race tests, vet and build; and swift build --build-tests all passed. Component/MyGo freshness checks and git diff --check passed. Current-head remote CI for merge commit 790ca9e0e6a807368a7b61db49021ed11d178f58 is complete: 16 checks passed, 0 failed, 0 pending, 0 skipped, including Apple, Android and all three MyGo platforms. This is the new merge head's result, not the earlier head's CI.

The original single shared contract case covers kind/value/profile admission; an existing native property-removal scenario proves default shrink -> factor 0 -> removal and basis preservation, with a red run when the new native mapping is absent. No additional tests, helpers or fixtures were introduced for conflict resolution.

Historical Swift caveat: the original local c6f18c26 full-suite attempt had one unresolved fullWidthIdentifiers focus/pinned-row failure (212 tests, one issue); its focused base run passed, so that attempt did not prove a baseline failure. The suite includes AppKit windows and should not have been run under this task's no-GUI limit; further local window tests were stopped. Subsequent pure retained sparse-contract tests passed 8/8 on that head and exact base, but do not replace GUI evidence. The merge includes main's separate sparse-focus recovery fix; this conflict-resolution task only compiled Swift tests and did not rerun AppKit or establish the original failure's cause. Application GUI acceptance was not performed.

Link to Devin session: https://app.devin.ai/sessions/eccad85c243745debb033bc62ba68c4e
Open in Devin Desktop: https://app.devin.ai/desktop/session/eccad85c243745debb033bc62ba68c4e?variant=devin
Requested by: @tiensonqin

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring