Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →AI エージェントのアクション空間、ツール定義、観測フォーマットを設計・最適化して完了率を向上させます。
.claude/skills/affaan-m-agent-harness-construction/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 11% | 0% |
当你在改进智能体的规划、调用工具、从错误中恢复以及收敛到完成状态的方式时，使用此技能。
智能体输出质量受限于：
每个工具响应都应包括：
status: success|warning|errorsummary: 一行结果next_actions: 可执行的后续步骤artifacts: 文件路径 / ID对于每个错误路径，应包括：
跟踪：
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 15,440 | 10,912 | -29% | 1 | 1 | 0% | 2,489 | 2,343 | -6% | 0 | 0 | — |
case-01 | fail→pass | 21,875 | 18,478 | -16% | 1 | 1 | 0% | 3,553 | 4,011 | +13% | 0 | 0 | — |
case-02 | fail→pass | 23,828 | 21,023 | -12% | 1 | 1 | 0% | 3,998 | 4,158 | +4% | 0 | 0 | — |
case-03 | fail→pass | 25,124 | 23,435 | -7% | 1 | 1 | 0% | 4,201 | 4,646 | +11% | 0 | 0 | — |
case-04 | fail→pass | 14,374 | 9,211 | -36% | 1 | 1 | 0% | 2,870 | 2,370 | -17% | 0 | 0 | — |
case-05 | pass→pass | 12,761 | 10,326 | -19% | 1 | 1 | 0% | 2,209 | 2,370 | +7% | 0 | 0 | — |
case-07 | pass→pass | 14,234 | 12,362 | -13% | 1 | 1 | 0% | 2,897 | 2,973 | +3% | 0 | 0 | — |
case-08 | pass→pass | 15,574 | 11,487 | -26% | 1 | 1 | 0% | 2,848 | 2,502 | -12% | 0 | 0 | — |
case-09 | fail→pass | 14,594 | 12,692 | -13% | 1 | 1 | 0% | 2,610 | 2,901 | +11% | 0 | 0 | — |
case-10 | fail→pass | 8,016 | 3,984 | -50% | 1 | 1 | 0% | 1,336 | 1,156 | -13% | 0 | 0 | — |
case-11 | pass→pass | 14,699 | 13,705 | -7% | 1 | 1 | 0% | 2,639 | 2,852 | +8% | 0 | 0 | — |
case-12 | pass→pass | 9,181 | 6,439 | -30% | 1 | 1 | 0% | 1,485 | 1,542 | +4% | 0 | 0 | — |
case-13 | pass→pass | 8,793 | 5,821 | -34% | 1 | 1 | 0% | 1,557 | 1,667 | +7% | 0 | 0 | — |
case-14 | pass→pass | 12,125 | 3,079 | -75% | 1 | 1 | 0% | 2,096 | 1,017 | -51% | 0 | 0 | — |
case-15 | pass→pass | 14,574 | 9,643 | -34% | 1 | 1 | 0% | 2,437 | 2,107 | -14% | 0 | 0 | — |
case-22 | pass→pass | 15,360 | 12,603 | -18% | 1 | 1 | 0% | 2,891 | 3,065 | +6% | 0 | 0 | — |
case-16 | pass→pass | 12,815 | 11,307 | -12% | 1 | 1 | 0% | 2,123 | 2,453 | +16% | 0 | 0 | — |
case-17 | pass→pass | 9,404 | 6,744 | -28% | 1 | 1 | 0% | 1,766 | 1,633 | -8% | 0 | 0 | — |
case-18 | pass→pass | 16,974 | 6,819 | -60% | 1 | 1 | 0% | 2,660 | 1,686 | -37% | 0 | 0 | — |
case-19 | fail→pass | 13,719 | 12,861 | -6% | 1 | 1 | 0% | 2,317 | 2,495 | +8% | 0 | 0 | — |
case-20 | pass→pass | 11,436 | 11,043 | -3% | 1 | 1 | 0% | 2,285 | 2,795 | +22% | 0 | 0 | — |
case-21 | pass→pass | 13,920 | 12,081 | -13% | 1 | 1 | 0% | 2,641 | 2,933 | +11% | 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 +32 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.