Install any skill in seconds. Free to start, no credit card required.
Get Started Free →基于解析后的文档与用户意图,生成四色卡片——事实蓝卡、解释绿卡、风险黄卡、行动红卡,是八官署方法论的核心分析输出。
.claude/skills/anbeime-antinet-four-color-cards/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -31% | 0% |
markdown:密卷房解析产出的结构化文档intent:用户意图 / 查询目标llm_available:LLM 是否可用(影响黄卡生成路径)blue_card:事实卡——可溯源的陈述,附来源片段green_card:解释卡——对事实的机制/背景说明yellow_card:风险卡——过度声明、不确定性与潜在偏差(经监察院复核)red_card:行动卡——下一步建议(经丞相府补全)pending_review。scripts/run_four_color_cards.py --stage {extract|review|propose}extract → 通政司 comm.tongzhengsi.TongZhengSiAgent(蓝卡:事实抽取)review → 监察院 audit.jichayuan.JianChaYuanAgent(绿卡:Gap/解释)propose → 丞相府 strategy.chengxiangfu.ChengXiangFuAgent(红卡:行动建议)python skills/four-color-cards/scripts/run_four_color_cards.py --stage extractpython skills/four-color-cards/scripts/run_four_color_cards.py --stage reviewpython skills/four-color-cards/scripts/run_four_color_cards.py --stage proposeexamples/snse_survey/skill_outputs/four_color_<stage>.json| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 24,469 | 31,441 | +28% | 1 | 1 | 0% | 2,827 | 5,691 | +101% | 0 | 0 | — |
case-02 | pass→pass | 25,058 | 8,585 | -66% | 1 | 1 | 0% | 3,037 | 1,407 | -54% | 0 | 0 | — |
case-03 | fail→pass | 21,693 | 8,367 | -61% | 1 | 1 | 0% | 2,859 | 1,369 | -52% | 0 | 0 | — |
case-04 | fail→pass | 38,120 | 8,133 | -79% | 1 | 1 | 0% | 1,812 | 1,329 | -27% | 0 | 0 | — |
case-05 | fail→pass | 17,879 | 8,491 | -53% | 1 | 1 | 0% | 2,146 | 1,414 | -34% | 0 | 0 | — |
case-06 | fail→pass | 30,904 | 8,850 | -71% | 1 | 1 | 0% | 1,669 | 1,463 | -12% | 0 | 0 | — |
case-07 | fail→pass | 15,548 | 7,492 | -52% | 1 | 1 | 0% | 1,670 | 1,157 | -31% | 0 | 0 | — |
case-08 | fail→pass | 13,970 | 10,748 | -23% | 1 | 1 | 0% | 1,500 | 1,728 | +15% | 0 | 0 | — |
case-09 | fail→pass | 14,581 | 10,185 | -30% | 1 | 1 | 0% | 1,453 | 1,546 | +6% | 0 | 0 | — |
case-10 | fail→pass | 13,592 | 11,206 | -18% | 1 | 1 | 0% | 1,258 | 1,764 | +40% | 0 | 0 | — |
case-11 | fail→pass | 22,351 | 10,996 | -51% | 1 | 1 | 0% | 2,907 | 1,803 | -38% | 0 | 0 | — |
case-12 | fail→pass | 20,621 | 6,978 | -66% | 1 | 1 | 0% | 2,661 | 1,135 | -57% | 0 | 0 | — |
case-13 | fail→pass | 13,523 | 7,431 | -45% | 1 | 1 | 0% | 1,405 | 1,127 | -20% | 0 | 0 | — |
case-14 | fail→pass | 15,864 | 7,027 | -56% | 1 | 1 | 0% | 1,785 | 1,032 | -42% | 0 | 0 | — |
case-15 | fail→pass | 17,975 | 6,706 | -63% | 1 | 1 | 0% | 2,279 | 1,082 | -53% | 0 | 0 | — |
case-16 | fail→pass | 21,774 | 7,728 | -65% | 1 | 1 | 0% | 2,575 | 1,214 | -53% | 0 | 0 | — |
case-17 | fail→pass | 16,526 | 6,907 | -58% | 1 | 1 | 0% | 1,921 | 1,061 | -45% | 0 | 0 | — |
case-18 | fail→pass | 13,990 | 7,616 | -46% | 1 | 1 | 0% | 1,311 | 1,167 | -11% | 0 | 0 | — |
case-19 | fail→pass | 22,153 | 7,903 | -64% | 1 | 1 | 0% | 2,674 | 1,215 | -55% | 0 | 0 | — |
case-20 | fail→pass | 19,735 | 8,936 | -55% | 1 | 1 | 0% | 2,289 | 1,430 | -38% | 0 | 0 | — |
case-21 | fail→pass | 31,394 | 8,725 | -72% | 1 | 1 | 0% | 4,266 | 1,318 | -69% | 0 | 0 | — |
case-22 | fail→pass | 14,500 | 7,770 | -46% | 1 | 1 | 0% | 1,406 | 1,255 | -11% | 0 | 0 | — |
case-23 | fail→pass | 24,024 | 7,806 | -68% | 1 | 1 | 0% | 3,188 | 1,268 | -60% | 0 | 0 | — |
case-24 | fail→pass | 21,527 | 7,520 | -65% | 1 | 1 | 0% | 2,931 | 1,219 | -58% | 0 | 0 | — |
case-25 | fail→pass | 21,145 | 7,526 | -64% | 1 | 1 | 0% | 2,671 | 1,187 | -56% | 0 | 0 | — |
case-26 | pass→pass | 15,757 | 10,837 | -31% | 1 | 1 | 0% | 1,680 | 1,649 | -2% | 0 | 0 | — |
case-27 | pass→pass | 20,878 | 10,086 | -52% | 1 | 1 | 0% | 2,261 | 1,656 | -27% | 0 | 0 | — |
case-28 | pass→pass | 20,707 | 24,454 | +18% | 1 | 1 | 0% | 3,031 | 4,501 | +48% | 0 | 0 | — |
case-29 | pass→pass | 19,502 | 18,652 | -4% | 1 | 1 | 0% | 2,647 | 3,150 | +19% | 0 | 0 | — |
case-30 | pass→pass | 21,984 | 21,091 | -4% | 1 | 1 | 0% | 2,724 | 3,391 | +24% | 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. 30 cases were attempted, and 29 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 +77 percentage points is the difference between those two pass rates over the 29 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.