Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Decision needed + recommendation + evidence + tradeoffs, 把复杂材料压成可拍板的一页
.claude/skills/nexu-io-exec-briefing-memo/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 340% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 340% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 201% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 276% | 0% |
【模板: 高管决策简报 / Executive Briefing Memo】 【意图】这不是会议纪要、不是周报、不是 PRD。它的唯一目标是帮助决策者在 3 分钟内理解问题并拍板。
【适合输入】
【必须输出的结构】
【设计要求】
【可选风格模板 — 参考 assets/】 根据决策场景选择一种, 不要三种混用:
assets/board-memo.html: 默认风格。浅色高管 memo, 适合 CEO/CFO/CRO、运营、产品决策。assets/decision-command.html: 深色 command center, 适合紧急决策、风险处置、incident、go/no-go、launch gate。assets/board-paper.html: 正式 board paper / 董事会纸质议案, 适合董事会、投资人、合规、预算审批。如果用户没有指定风格, 优先使用 board-memo; 如果材料强调紧急、风险、行动指挥, 使用 decision-command; 如果材料面向董事会或正式审批, 使用 board-paper。
【内容真实性】
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 18,833 | 24,331 | +29% | 1 | 1 | 0% | 2,778 | 4,183 | +51% | 0 | 0 | — |
case-24 | pass→pass | 11,983 | 28,662 | +139% | 1 | 1 | 0% | 1,846 | 6,756 | +266% | 0 | 0 | — |
case-11 | fail→pass | 10,406 | 30,578 | +194% | 1 | 1 | 0% | 1,538 | 6,760 | +340% | 0 | 0 | — |
case-17 | pass→pass | 12,512 | 27,513 | +120% | 1 | 1 | 0% | 2,063 | 6,763 | +228% | 0 | 0 | — |
case-18 | fail→fail | 12,031 | 26,656 | +122% | 1 | 1 | 0% | 2,132 | 6,770 | +218% | 0 | 0 | — |
case-02 | fail→fail | 20,900 | 33,386 | +60% | 1 | 1 | 0% | 2,986 | 6,818 | +128% | 0 | 0 | — |
case-03 | fail→fail | 21,656 | 30,959 | +43% | 1 | 1 | 0% | 3,334 | 6,818 | +104% | 0 | 0 | — |
case-04 | fail→fail | 2,746 | 21,190 | +672% | 1 | 1 | 0% | 380 | 4,222 | +1011% | 0 | 0 | — |
case-05 | fail→fail | 4,895 | 21,373 | +337% | 1 | 1 | 0% | 768 | 4,166 | +442% | 0 | 0 | — |
case-06 | pass→fail | 22,907 | 30,502 | +33% | 1 | 1 | 0% | 3,659 | 6,775 | +85% | 0 | 0 | — |
case-07 | fail→pass | 10,196 | 28,633 | +181% | 1 | 1 | 0% | 1,537 | 6,765 | +340% | 0 | 0 | — |
case-08 | pass→pass | 12,157 | 29,218 | +140% | 1 | 1 | 0% | 1,684 | 6,666 | +296% | 0 | 0 | — |
case-09 | pass→pass | 13,071 | 29,421 | +125% | 1 | 1 | 0% | 2,014 | 6,758 | +236% | 0 | 0 | — |
case-10 | pass→pass | 13,614 | 47,850 | +251% | 1 | 1 | 0% | 1,862 | 6,753 | +263% | 0 | 0 | — |
case-12 | pass→pass | 9,164 | 29,964 | +227% | 1 | 1 | 0% | 1,265 | 6,751 | +434% | 0 | 0 | — |
case-13 | fail→fail | 16,965 | 38,786 | +129% | 1 | 1 | 0% | 2,902 | 6,758 | +133% | 0 | 0 | — |
case-14 | fail→pass | 14,553 | 30,352 | +109% | 1 | 1 | 0% | 2,243 | 6,752 | +201% | 0 | 0 | — |
case-15 | pass→fail | 18,951 | 25,882 | +37% | 1 | 1 | 0% | 3,330 | 6,757 | +103% | 0 | 0 | — |
case-16 | fail→fail | 9,741 | 25,354 | +160% | 1 | 1 | 0% | 1,632 | 6,755 | +314% | 0 | 0 | — |
case-19 | fail→fail | 10,101 | 17,035 | +69% | 1 | 1 | 0% | 1,433 | 3,602 | +151% | 0 | 0 | — |
case-20 | fail→fail | 12,706 | 27,238 | +114% | 1 | 1 | 0% | 2,144 | 6,769 | +216% | 0 | 0 | — |
case-21 | fail→pass | 12,602 | 28,882 | +129% | 1 | 1 | 0% | 1,797 | 6,763 | +276% | 0 | 0 | — |
case-22 | pass→pass | 18,781 | 26,489 | +41% | 1 | 1 | 0% | 2,878 | 6,360 | +121% | 0 | 0 | — |
case-23 | pass→pass | 12,848 | 26,590 | +107% | 1 | 1 | 0% | 2,064 | 6,760 | +228% | 0 | 0 | — |
case-25 | fail→fail | 15,585 | 27,446 | +76% | 1 | 1 | 0% | 2,667 | 6,755 | +153% | 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. 25 cases were attempted. The headline lift of +12 percentage points is the difference between those two pass rates over the 25 comparable cases. 2 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.