Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →オブザーバビリティ、セキュリティ境界、およびライフサイクル管理を備えた長寿命エージェントワークロードを運用します。
.claude/skills/affaan-m-enterprise-agent-ops/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -6% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -1% | 0% |
使用此技能用于需要超越单次 CLI 会话操作控制的云托管或持续运行的智能体系统。
当故障激增时：
此技能可与以下工具配合使用：
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 21,756 | 20,636 | -5% | 1 | 1 | 0% | 3,494 | 3,661 | +5% | 0 | 0 | — |
case-02 | fail→fail | 20,617 | 23,938 | +16% | 1 | 1 | 0% | 3,214 | 3,940 | +23% | 0 | 0 | — |
case-03 | fail→pass | 27,256 | 48,245 | +77% | 1 | 1 | 0% | 4,172 | 3,919 | -6% | 0 | 0 | — |
case-04 | fail→pass | 19,340 | 17,684 | -9% | 1 | 1 | 0% | 2,964 | 2,922 | -1% | 0 | 0 | — |
case-05 | pass→pass | 14,926 | 16,027 | +7% | 1 | 1 | 0% | 2,278 | 2,518 | +11% | 0 | 0 | — |
case-06 | fail→pass | 21,421 | 25,019 | +17% | 1 | 1 | 0% | 3,163 | 2,910 | -8% | 0 | 0 | — |
case-07 | pass→pass | 17,773 | 9,942 | -44% | 1 | 1 | 0% | 2,738 | 1,817 | -34% | 0 | 0 | — |
case-08 | fail→pass | 21,197 | 17,285 | -18% | 1 | 1 | 0% | 3,018 | 3,001 | -1% | 0 | 0 | — |
case-09 | fail→pass | 16,036 | 14,373 | -10% | 1 | 1 | 0% | 2,518 | 2,660 | +6% | 0 | 0 | — |
case-10 | pass→pass | 13,405 | 12,618 | -6% | 1 | 1 | 0% | 2,269 | 2,484 | +9% | 0 | 0 | — |
case-11 | pass→pass | 16,431 | 13,472 | -18% | 1 | 1 | 0% | 2,478 | 2,375 | -4% | 0 | 0 | — |
case-12 | fail→pass | 13,575 | 8,618 | -37% | 1 | 1 | 0% | 1,749 | 1,502 | -14% | 0 | 0 | — |
case-13 | pass→pass | 21,522 | 15,053 | -30% | 1 | 1 | 0% | 2,855 | 2,500 | -12% | 0 | 0 | — |
case-14 | pass→pass | 19,085 | 13,078 | -31% | 1 | 1 | 0% | 2,660 | 2,153 | -19% | 0 | 0 | — |
case-15 | fail→pass | 17,980 | 13,096 | -27% | 1 | 1 | 0% | 2,715 | 2,275 | -16% | 0 | 0 | — |
case-16 | fail→pass | 14,110 | 4,691 | -67% | 1 | 1 | 0% | 2,203 | 1,000 | -55% | 0 | 0 | — |
case-17 | pass→pass | 16,165 | 22,471 | +39% | 1 | 1 | 0% | 2,446 | 2,508 | +3% | 0 | 0 | — |
case-18 | pass→pass | 19,368 | 17,556 | -9% | 1 | 1 | 0% | 2,897 | 3,050 | +5% | 0 | 0 | — |
case-19 | pass→pass | 16,908 | 9,605 | -43% | 1 | 1 | 0% | 2,462 | 1,793 | -27% | 0 | 0 | — |
case-20 | fail→pass | 17,599 | 15,566 | -12% | 1 | 1 | 0% | 2,523 | 2,482 | -2% | 0 | 0 | — |
case-21 | pass→pass | 18,282 | 16,265 | -11% | 1 | 1 | 0% | 2,959 | 3,162 | +7% | 0 | 0 | — |
case-22 | pass→pass | 16,851 | 29,243 | +74% | 1 | 1 | 0% | 2,911 | 3,511 | +21% | 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. The headline lift of +45 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.
Other measured skills in the registry, with their headline benchmark lift.