Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when generated code looks wrong, code generation fails, or you need to understand FRB internals
.claude/skills/fzyzcjy-frb-debugging/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -66% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -38% | 0% |
> Note: Read frb-dev-env before running commands and follow the active user's environment rules.
The in-tree examples by default enable dump_all: true in flutter_rust_bridge.yaml for debugability.
Therefore, you can check rust/target/frb_dump/ for:
bashRUST_LOG=debug flutter_rust_bridge_codegen ...
If above doesn't help, see:
website/docs/docs/guides/contributing/tip.mdwebsite/docs/manual/troubleshooting.mdBefore going deep into FRB internals, first check whether you are actually seeing failure propagation from unstable generated outputs or integrate templates.
In particular:
frb-fix-ci firstfrb-debugging after frb-fix-ci has ruled out those simpler explanationsfrb-code-generation - Which generation commands to runfrb-fix-ci - Diagnose CI failure propagation before deep debuggingfrb-develop-feature - Development workflow| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 20,241 | 16,419 | -19% | 1 | 1 | 0% | 3,137 | 2,006 | -36% | 0 | 0 | — |
case-02 | fail→fail | 18,675 | 14,361 | -23% | 1 | 1 | 0% | 2,633 | 1,669 | -37% | 0 | 0 | — |
case-03 | fail→pass | 22,523 | 10,304 | -54% | 1 | 1 | 0% | 2,605 | 2,002 | -23% | 0 | 0 | — |
case-04 | fail→pass | 12,932 | 10,376 | -20% | 1 | 1 | 0% | 1,936 | 1,138 | -41% | 0 | 0 | — |
case-05 | fail→pass | 16,821 | 8,436 | -50% | 1 | 1 | 0% | 1,618 | 551 | -66% | 0 | 0 | — |
case-06 | fail→fail | 10,610 | 9,888 | -7% | 1 | 1 | 0% | 1,680 | 767 | -54% | 0 | 0 | — |
case-07 | pass→pass | 27,725 | 10,429 | -62% | 1 | 1 | 0% | 1,344 | 864 | -36% | 0 | 0 | — |
case-08 | fail→fail | 19,758 | 2,844 | -86% | 1 | 1 | 0% | 2,485 | 671 | -73% | 0 | 0 | — |
case-09 | fail→fail | 19,586 | 7,436 | -62% | 1 | 1 | 0% | 2,667 | 580 | -78% | 0 | 0 | — |
case-10 | fail→pass | 22,632 | 14,985 | -34% | 1 | 1 | 0% | 2,944 | 1,814 | -38% | 0 | 0 | — |
case-11 | fail→pass | 9,681 | 7,343 | -24% | 1 | 1 | 0% | 1,598 | 623 | -61% | 0 | 0 | — |
case-12 | fail→pass | 16,318 | 8,760 | -46% | 1 | 1 | 0% | 2,495 | 756 | -70% | 0 | 0 | — |
case-13 | fail→pass | 14,243 | 3,176 | -78% | 1 | 1 | 0% | 1,385 | 697 | -50% | 0 | 0 | — |
case-14 | pass→pass | 7,919 | 8,222 | +4% | 1 | 1 | 0% | 1,241 | 837 | -33% | 0 | 0 | — |
case-15 | fail→pass | 26,550 | 9,480 | -64% | 1 | 1 | 0% | 3,172 | 989 | -69% | 0 | 0 | — |
case-16 | pass→pass | 18,196 | 7,810 | -57% | 1 | 1 | 0% | 2,100 | 670 | -68% | 0 | 0 | — |
case-17 | pass→pass | 20,335 | 8,737 | -57% | 1 | 1 | 0% | 2,077 | 789 | -62% | 0 | 0 | — |
case-18 | fail→pass | 17,630 | 2,905 | -84% | 1 | 1 | 0% | 1,879 | 696 | -63% | 0 | 0 | — |
case-19 | fail→pass | 16,578 | 2,828 | -83% | 1 | 1 | 0% | 1,944 | 703 | -64% | 0 | 0 | — |
case-20 | fail→pass | 30,276 | 8,151 | -73% | 1 | 1 | 0% | 1,177 | 664 | -44% | 0 | 0 | — |
case-21 | pass→pass | 13,812 | 10,525 | -24% | 1 | 1 | 0% | 2,581 | 1,827 | -29% | 0 | 0 | — |
case-22 | pass→pass | 21,627 | 15,665 | -28% | 1 | 1 | 0% | 2,795 | 2,941 | +5% | 0 | 0 | — |
case-23 | pass→pass | 12,603 | 10,390 | -18% | 1 | 1 | 0% | 1,190 | 1,083 | -9% | 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, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +52 percentage points is the difference between those two pass rates over the 22 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.