Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Upgrade flutter_rust_bridge to a new Flutter stable release. Use when changing Flutter/Dart versions, devcontainer Docker images, CI/post-release pins, generated Flutter scaffolds, or platform compatibility.
.claude/skills/fzyzcjy-frb-upgrade-flutter/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 28% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 16% | 0% |
frb-dev-env before running setup, generation, lint, or tests. Use its per-worktree local Docker workflow.frb-upgrade-docker when the target toolchain changes the development image.frb-docker for ordinary local Docker usage.frb-code-generation before accepting generated or scaffold drift.frb-cargokit before changing any copied CargoKit file. Read frb-cargokit-dev only when the sourcechange belongs in the external CargoKit repository.
frb-pr-chain-split as soon as an independently landable prerequisite or cleanup appears.frb-prepare-pr and then frb-pr-review before treating the upgrade PR as ready.frb-fix-ci when CI starts failing.shellrg -n "FRB_MAIN_|FLUTTER_VERSION|DART_VERSION|RUST_VERSION|setup-flutter|setup-dart|cirruslabs/flutter" rg -n "flutter_rust_bridge_dev|stable|nightly|minimum.*version|deployment.*target" \ .devcontainer .github tools frb_codegen frb_example
.devcontainer/Dockerfile;.github/workflows/ci.yaml and .github/workflows/post_release.yaml;.github/workflows/publish_dev_docker.yaml;tools/frb_internal/test/src/makefile_dart/test_dev_docker_metadata.dart;pubspec.yaml and checked-in pubspec.lock files;frb_codegen/assets/integration_template/**;tools/frb_internal/assets/apple_scaffold/**;tools/tart_macos/**;frb_example/**.master.Follow frb-upgrade-docker for its contents, candidate image, merge order, and stable-tag promotion.
the main upgrade PR.
frb-pr-chain-split.
gh stack workflow described there.ci-manual-dispatch on dormant predecessors unless a predecessor specifically needs GitHub-only validation.master.frb-upgrade-docker before relying on the upgraded image.latest from a PR branch.master into the remaining upgrade chain anduse the canonical version tag.
frb-dev-env andread frb-tart-prepare before provisioning or validation.
.github/workflows/ci.yaml and.github/workflows/post_release.yaml:
FRB_MAIN_FLUTTER_VERSION;FRB_MAIN_DART_VERSION;FRB_MAIN_RUST_VERSION when the upgraded tooling requires it;FRB_RUSTFMT_NIGHTLY_VERSION only when formatting or nightly rust-src behavior requires it.that compatibility lane.
assumptions.
frb-code-generation for command selection, convergence, generated-output provenance, legacy scaffold migrations, and OHOS integrate composition.frb-cargokit and frb-cargokit-dev for CargoKit ownership and synchronization.frb-dev-env. When the Docker image changed, use the candidate or canonical imageselected by frb-upgrade-docker; do not reuse a stale per-worktree container.
frb-lint and frb-test for exact commands.frb-prepare-pr.frb-pr-review.frb-fix-ci before deep CI debugging.then coverage or uploads.
frb-pr-chain-split for filtered or deferred predecessor CI.frb-upgrade-docker; do not defer imagepublication until the main Flutter upgrade PR merges.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 22,420 | 32,978 | +47% | 1 | 1 | 0% | 2,949 | 4,448 | +51% | 0 | 0 | — |
case-02 | fail→pass | 22,883 | 16,708 | -27% | 1 | 1 | 0% | 3,105 | 3,677 | +18% | 0 | 0 | — |
case-03 | fail→pass | 26,291 | 24,627 | -6% | 1 | 1 | 0% | 3,649 | 4,414 | +21% | 0 | 0 | — |
case-04 | pass→pass | 20,057 | 12,477 | -38% | 1 | 1 | 0% | 2,545 | 2,679 | +5% | 0 | 0 | — |
case-05 | fail→pass | 17,447 | 11,134 | -36% | 1 | 1 | 0% | 2,147 | 2,751 | +28% | 0 | 0 | — |
case-06 | pass→pass | 19,523 | 10,983 | -44% | 1 | 1 | 0% | 2,304 | 2,729 | +18% | 0 | 0 | — |
case-07 | fail→pass | 17,251 | 10,784 | -37% | 1 | 1 | 0% | 2,040 | 2,372 | +16% | 0 | 0 | — |
case-08 | fail→pass | 20,001 | 9,487 | -53% | 1 | 1 | 0% | 2,627 | 2,467 | -6% | 0 | 0 | — |
case-09 | fail→pass | 14,022 | 10,878 | -22% | 1 | 1 | 0% | 1,539 | 2,642 | +72% | 0 | 0 | — |
case-10 | fail→pass | 16,002 | 10,056 | -37% | 1 | 1 | 0% | 1,662 | 2,331 | +40% | 0 | 0 | — |
case-11 | fail→pass | 17,627 | 7,249 | -59% | 1 | 1 | 0% | 1,773 | 3,083 | +74% | 0 | 0 | — |
case-12 | fail→pass | 20,574 | 14,329 | -30% | 1 | 1 | 0% | 2,440 | 3,123 | +28% | 0 | 0 | — |
case-13 | fail→pass | 22,593 | 33,611 | +49% | 1 | 1 | 0% | 2,753 | 2,935 | +7% | 0 | 0 | — |
case-14 | fail→pass | 14,929 | 10,079 | -32% | 1 | 1 | 0% | 1,632 | 2,467 | +51% | 0 | 0 | — |
case-15 | fail→pass | 11,073 | 10,909 | -1% | 1 | 1 | 0% | 1,788 | 2,423 | +36% | 0 | 0 | — |
case-16 | pass→pass | 8,847 | 9,391 | +6% | 1 | 1 | 0% | 1,378 | 2,376 | +72% | 0 | 0 | — |
case-17 | fail→pass | 11,934 | 8,619 | -28% | 1 | 1 | 0% | 1,778 | 1,944 | +9% | 0 | 0 | — |
case-18 | fail→pass | 14,681 | 4,595 | -69% | 1 | 1 | 0% | 2,427 | 2,303 | -5% | 0 | 0 | — |
case-19 | pass→pass | 22,407 | 11,828 | -47% | 1 | 1 | 0% | 3,168 | 2,561 | -19% | 0 | 0 | — |
case-20 | pass→pass | 17,396 | 9,283 | -47% | 1 | 1 | 0% | 2,903 | 3,089 | +6% | 0 | 0 | — |
case-21 | pass→pass | 16,906 | 8,831 | -48% | 1 | 1 | 0% | 3,230 | 3,238 | +0% | 0 | 0 | — |
case-22 | pass→pass | 17,918 | 15,715 | -12% | 1 | 1 | 0% | 2,219 | 3,597 | +62% | 0 | 0 | — |
case-23 | pass→pass | 5,670 | 4,037 | -29% | 1 | 1 | 0% | 1,038 | 2,272 | +119% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +65 percentage points is the difference between those two pass rates over the 23 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/9/2026 | +59% |
Other measured skills in the registry, with their headline benchmark lift.