chore(release): 0.0.10

[anilcancakir]

Cuts 0.0.10 so MagicSelector is reachable from a consumer. watchools needs it to scope the rebuild a keystroke causes, and .claude/rules/workflow.md there forbids reshaping the app around an unreleased API: its CI resolves magic from pub.dev, so wiring the selector in before this lands would go red on a green local run.

What ships

The whole [Unreleased] section, promoted unchanged. Four groups:

• BREAKING: file_picker moves from >=11.0.2 <12.0.0-0 to ^12.2.0 and the Pick facade is ported to v12's surface (pickFiles returns a plain list, saveFile returns a Uri?, withData is gone). Pre-1.0, so a breaking change in a patch is what CLAUDE.md allows.
• Added: MagicSelector<C, T>, plus the earlier entries already sitting there.
• Changed and Fixed: unchanged from what was already recorded.

The four surfaces the checklist names

Surface	Change
pubspec.yaml	version: 0.0.9 to 0.0.10
CHANGELOG.md	## [Unreleased] to ## [0.0.10] - 2026-09-09
skills/magic-framework/SKILL.md	stamp comment magic 0.0.9 to magic 0.0.10
example/pubspec.yaml	pinned version: 0.0.9 to 0.0.10

No trailing link references in this CHANGELOG, so there is nothing to add beside the heading.

Gates

dart analyze clean, dart format . no diff, 1440 tests green with no new skips, dart pub publish --dry-run clean apart from the expected uncommitted-files warning at the time it ran.

One hint from the dry run worth passing on rather than silently accepting: pub.dev's newest version of this package is 1.0.0-alpha.13, so 0.0.10 is behind it on the prerelease line while being ahead on the stable one. CLAUDE.md declares master the active 0.0.x line and 1.0.0 the future milestone, so this is consistent with the stated plan, but the two facts sit oddly together and someone should decide whether the alpha line is still meant to exist.

After this merges

git tag 0.0.10 && git push origin 0.0.10, which is what publish.yml watches. I have not tagged anything: the publish is irreversible and outward-facing, so it is the maintainer's call rather than mine.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[kodizm]

Note

Kodizm (AI-generated). May contain mistakes; verify before acting.

The four release surfaces are all correct and consistent, and the release notes will extract cleanly; two documentation gaps in the promoted section, neither blocking.

Minor

CHANGELOG.md:5 — the heading is dated 2026-09-09, which is tomorrow (today is 2026-09-08). Harmless if the merge and tag land tomorrow; otherwise the published GitHub release and the pub.dev changelog carry a date the release did not happen on. Note the SKILL.md stamp beside it still reads (2026-09-08), so the two disagree by a day either way. (maintainability)

pubspec.yaml:31 — the promoted section documents the file_picker constraint move as BREAKING but says nothing about go_router, which also changed since the last release and ships here. 0.0.9 pinned ^17.1.0; HEAD carries >=17.1.0 <19.0.0 (widened in #144, which git merge-base --is-ancestor 3a34e0f 0.0.9 confirms is not in the 0.0.9 tag). Widening a range is not itself breaking, but lib/magic.dart:6 re-exports package:go_router/go_router.dart wholesale, and a fresh flutter pub get here resolves go_router 18.0.1 — so go_router 18's breaking changes reach consumers through magic's barrel on a patch bump with no changelog line pointing at it. A ### Changed entry would be enough. (maintainability)

Notes, not findings

• The 1.0.0-alpha.13 dry-run hint you flagged is consistent with CLAUDE.md, which declares master the active 0.0.x line and 1.0.0 the future milestone. Agreed it is worth a separate decision rather than silence, and agreed it does not block this.
• I verified the release-notes extraction rather than assuming it: publish.yml's awk against ## [0.0.10] returns the 31-line section and stops correctly at ## [0.0.9].
• Every feature the section claims exists at HEAD: MagicSelector at lib/src/ui/magic_selector.dart, exported at lib/magic.dart:91, documented in doc/basics/ui-helpers.md:355 and skills/magic-framework/references/controllers-views.md:403, and file_picker: ^12.2.0 at pubspec.yaml:47. No stale 0.0.9 reference remains outside the changelog history (the two grep hits are fluttersdk_artisan and fluttersdk_dusk, unrelated).

Tests

Not a code change, so nothing new to cover. I ran the suites for the features being released rather than trusting the section: 26 pass in magic_selector_test.dart + magic_paginated_list_view_test.dart (including a refresh does not wear the loading-more footer, the named Fixed entry), and 60 pass in pick_test.dart + magic_paginator_test.dart.

Checks I ran

• flutter pub get - resolved; example/ resolved against the new version: 0.0.10 pin, so the path-plus-version pair is not stale.
• dart format --set-exit-if-changed . - clean, Formatted 336 files (0 changed).
• flutter test test/ui/magic_selector_test.dart test/ui/magic_paginated_list_view_test.dart - +26: All tests passed!
• flutter test test/facades/pick_test.dart test/http/magic_paginator_test.dart - +60: All tests passed!
• dart pub publish --dry-run - 1 warning + 1 hint; the warning is the analyzer crash below, not a code issue, and the hint is the alpha.13 one.
• flutter analyze / dart analyze lib bin - could not complete: the analysis server was killed (exited with code -9, then analysis server crashed unexpectedly) in this sandbox. That is a memory limit here, not a signal about the code, so I am not confirming or disputing your clean-analyze claim; CI will.
• Full flutter test (1440 tests) exceeded 20 minutes and I stopped it, so I did not observe the whole suite green - only the 86 tests above.