Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →你负责在 run 收尾后先判断是否需要沉淀知识对象，再生成对外交付报告与 session artifacts。`r`n`r`n你不参与当前 run 的前置方向建议，也不读取 triage 的知识匹配结果来反向做 dispatch。
.claude/skills/haolange-skill-report-knowledge-curator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -2% | 0% |
你负责在 run 收尾后先判断是否需要沉淀知识对象，再生成对外交付报告与 session artifacts。rnrn你不参与当前 run 的前置方向建议，也不读取 triage 的知识匹配结果来反向做 dispatch。
你只能在 rdc-debugger 明确 handoff 且 finalize 前置条件全部满足后工作；不得重判 intent gate，不得反向调度 specialist，不得补做前置 investigation。
进入 finalize 前必须同时满足：
fix_verification.yaml 有效skeptic_signoff.yaml strict passreports/report.md 必须存在reports/visual_report.html 必须存在任一项缺失时：
session_evidence.yamlreports/report.mdreports/visual_report.htmlreports/report.md 或 reports/visual_report.html 时宣告收尾完成| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-14 | pass→pass | 8,694 | 4,199 | -52% | 1 | 1 | 0% | 1,348 | 1,058 | -22% | 0 | 0 | — |
case-01 | fail→pass | 29,860 | 18,548 | -38% | 1 | 1 | 0% | 5,542 | 4,005 | -28% | 0 | 0 | — |
case-02 | fail→pass | 25,964 | 19,380 | -25% | 1 | 1 | 0% | 4,988 | 3,692 | -26% | 0 | 0 | — |
case-03 | fail→pass | 47,441 | 16,640 | -65% | 1 | 1 | 0% | 4,071 | 3,521 | -14% | 0 | 0 | — |
case-04 | pass→pass | 7,077 | 3,619 | -49% | 1 | 1 | 0% | 1,189 | 980 | -18% | 0 | 0 | — |
case-05 | pass→pass | 9,773 | 3,926 | -60% | 1 | 1 | 0% | 1,436 | 1,069 | -26% | 0 | 0 | — |
case-06 | pass→pass | 9,409 | 4,488 | -52% | 1 | 1 | 0% | 1,476 | 1,193 | -19% | 0 | 0 | — |
case-07 | fail→pass | 8,362 | 4,175 | -50% | 1 | 1 | 0% | 1,410 | 1,031 | -27% | 0 | 0 | — |
case-08 | fail→pass | 6,684 | 4,551 | -32% | 1 | 1 | 0% | 1,174 | 1,155 | -2% | 0 | 0 | — |
case-09 | fail→pass | 10,529 | 6,062 | -42% | 1 | 1 | 0% | 1,508 | 1,338 | -11% | 0 | 0 | — |
case-10 | fail→pass | 7,312 | 4,808 | -34% | 1 | 1 | 0% | 1,232 | 1,125 | -9% | 0 | 0 | — |
case-11 | pass→pass | 8,478 | 5,028 | -41% | 1 | 1 | 0% | 1,358 | 1,177 | -13% | 0 | 0 | — |
case-12 | fail→pass | 13,383 | 7,993 | -40% | 1 | 1 | 0% | 1,972 | 1,636 | -17% | 0 | 0 | — |
case-13 | fail→pass | 9,642 | 2,639 | -73% | 1 | 1 | 0% | 1,581 | 689 | -56% | 0 | 0 | — |
case-15 | fail→pass | 6,021 | 6,176 | +3% | 1 | 1 | 0% | 522 | 1,149 | +120% | 0 | 0 | — |
case-16 | fail→pass | 9,215 | 4,225 | -54% | 1 | 1 | 0% | 1,218 | 951 | -22% | 0 | 0 | — |
case-17 | pass→pass | 14,580 | 7,007 | -52% | 1 | 1 | 0% | 1,388 | 1,173 | -15% | 0 | 0 | — |
case-18 | pass→pass | 9,518 | 3,909 | -59% | 1 | 1 | 0% | 1,622 | 887 | -45% | 0 | 0 | — |
case-19 | fail→pass | 11,897 | 3,283 | -72% | 1 | 1 | 0% | 1,854 | 912 | -51% | 0 | 0 | — |
case-20 | fail→pass | 5,070 | 11,202 | +121% | 1 | 1 | 0% | 189 | 1,704 | +802% | 0 | 0 | — |
case-21 | fail→pass | 8,425 | 6,323 | -25% | 1 | 1 | 0% | 1,503 | 1,379 | -8% | 0 | 0 | — |
case-22 | fail→pass | 18,448 | 8,216 | -55% | 1 | 1 | 0% | 3,217 | 1,604 | -50% | 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 21 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 +68 percentage points is the difference between those two pass rates over the 21 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.