Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →CLI 자동화에서 기본 모델을 쓰다가 사용량 한도에 걸리면 다른 CLI로 자동 재실행하는 폴백을 구현한다.
.claude/skills/bam-bam-2-claude-codex-fallback/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 112% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 31% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 32% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 5% | 0% |
> 이 스킬이 시스템에 하는 일 (설치 전 확인) > > - claude와 codex CLI를 로컬에서 실행합니다. 둘 다 설치돼 있어야 합니다. > - OAuth 토큰을 ~/.config/llm/claude-oauth-token에서 읽기만 합니다. 출력하거나 전송하지 않습니다. > - 프롬프트는 임시파일(mktemp, mode 600)에 두고 종료 시 삭제합니다. > - 네트워크 호출은 각 CLI가 하는 것뿐이고, 이 스크립트가 따로 하는 외부 통신은 없습니다.
weekly limit, usage limit, hit your limit, limit ... resets| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 8,118 | 13,220 | +63% | 1 | 1 | 0% | 325 | 896 | +176% | 0 | 0 | — |
case-02 | fail→fail | 57,921 | 8,433 | -85% | 1 | 1 | 0% | 7,587 | 712 | -91% | 0 | 0 | — |
case-03 | fail→fail | 33,849 | 8,811 | -74% | 1 | 1 | 0% | 5,284 | 948 | -82% | 0 | 0 | — |
case-04 | pass→pass | 15,627 | 11,256 | -28% | 1 | 1 | 0% | 2,607 | 2,303 | -12% | 0 | 0 | — |
case-05 | pass→fail | 14,077 | 60,995 | +333% | 1 | 1 | 0% | 2,251 | 8,683 | +286% | 0 | 0 | — |
case-06 | fail→pass | 26,509 | 10,392 | -61% | 1 | 1 | 0% | 1,981 | 1,907 | -4% | 0 | 0 | — |
case-07 | fail→pass | 60,433 | 29,317 | -51% | 1 | 1 | 0% | 2,011 | 4,270 | +112% | 0 | 0 | — |
case-08 | fail→pass | 15,939 | 17,145 | +8% | 1 | 1 | 0% | 2,866 | 3,766 | +31% | 0 | 0 | — |
case-09 | fail→pass | 25,880 | 14,440 | -44% | 1 | 1 | 0% | 2,355 | 3,107 | +32% | 0 | 0 | — |
case-10 | fail→pass | 15,607 | 18,037 | +16% | 1 | 1 | 0% | 2,706 | 2,837 | +5% | 0 | 0 | — |
case-11 | fail→pass | 17,747 | 20,752 | +17% | 1 | 1 | 0% | 2,928 | 4,591 | +57% | 0 | 0 | — |
case-12 | fail→pass | 31,244 | 18,977 | -39% | 1 | 1 | 0% | 3,001 | 3,402 | +13% | 0 | 0 | — |
case-13 | fail→pass | 15,382 | 8,612 | -44% | 1 | 1 | 0% | 2,973 | 1,942 | -35% | 0 | 0 | — |
case-14 | fail→pass | 37,499 | 12,583 | -66% | 1 | 1 | 0% | 1,170 | 2,865 | +145% | 0 | 0 | — |
case-15 | pass→pass | 29,546 | 21,851 | -26% | 1 | 1 | 0% | 4,397 | 4,539 | +3% | 0 | 0 | — |
case-16 | fail→pass | 24,124 | 35,509 | +47% | 1 | 1 | 0% | 1,274 | 4,756 | +273% | 0 | 0 | — |
case-17 | pass→fail | 21,148 | 24,504 | +16% | 1 | 1 | 0% | 3,553 | 5,276 | +48% | 0 | 0 | — |
case-18 | fail→fail | 13,062 | 9,972 | -24% | 1 | 1 | 0% | 2,030 | 2,321 | +14% | 0 | 0 | — |
case-19 | fail→pass | 12,262 | 6,466 | -47% | 1 | 1 | 0% | 1,589 | 1,366 | -14% | 0 | 0 | — |
case-20 | pass→pass | 12,035 | 9,508 | -21% | 1 | 1 | 0% | 2,085 | 2,123 | +2% | 0 | 0 | — |
case-21 | fail→fail | 11,192 | 8,385 | -25% | 1 | 1 | 0% | 1,747 | 1,983 | +14% | 0 | 0 | — |
case-22 | pass→pass | 29,732 | 12,498 | -58% | 1 | 1 | 0% | 2,991 | 2,922 | -2% | 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. 22 cases were attempted, and 16 counted toward the lift figure. The other 6 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 +41 percentage points is the difference between those two pass rates over the 16 comparable cases. 3 cases got worse with the skill loaded, and they are included in that figure.
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.
Other measured skills in the registry, with their headline benchmark lift.