Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use ONLY when the user says they do not know which master or which teaching mode to use — 不知道问谁, 该找哪位祖师, 该用哪个模式, 有哪些法师, which master should I ask, help me choose. This is a router, not a teacher: it names a destination and stops. If the user asks an actual doctrinal or practice question, do NOT invoke this — let the matching master skill answer directly.
.claude/skills/xr843-master-help/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -56% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -20% | 0% |
> 本 skill 只做导航，不讲教义。选定目标后立即交棒，不要代替祖师回答。
用户不知道该用哪位祖师 / 哪个教学模式时，给出目标并停手。
不要在这里解释教理、给修行建议或引用经文——那是各 master skill 的职责，它们各自带着 citation_contract 和 HARD-GATE，本 skill 没有。
路由表在仓库根的 routing.json（mode_rules / topic_pairings / default_pairing）， 祖师关键词在各 prebuilt/<slug>/meta.json 的 search_scope.keywords。 两处都是机器可读的单一数据源——不要凭记忆列举祖师或关键词，读文件。
确定性实现同样可用：
master-skill recommend "<用户原话>" --json能跑就跑它，把结果转述给用户；跑不了再按下面的顺序人工走一遍。
与 routing.json.mode_rules 的 order 一致：
1. 命中「学习计划 / 入门 / 先学什么 / 从哪开始 / 按什么顺序」 → /master-curriculum
2. 命中「辩论 / 谁更对 / 高下 / 之争 / 之辩 / 分判」 → /master-debate
3. 命中「对比 / 比较 / 不同 / 异同 / 各派怎么看」 → /compare-masters
4. 都不命中 → 单位祖师：按 meta.json search_scope.keywords 打分
5. 仍无命中 → routing.json.situations 白话状况层
6. 仍无命中 → routing.json.topic_pairings 主题配对
7. 再无命中 → routing.json.default_pairing第 5 步是给说不出术语的人用的。search_scope.keywords 是教理检索词， 新手不会打"四念处"，他会打"坐不住"。用户描述的是感受（妄念 / 看不懂 / 无力感 / 想学最朴素的）而非主题时，走这一层。
第 4 步打分规则：关键词长度 ≥ 2 才计分（单字 空 戒 定 慧 苦 禅 业 会在日常汉语里误命中，已被 min_keyword_length 排除）；命中数高者优先； 平局时优先不同传统，仍平局按 slug 字典序。最多 3 位。
你的问题看起来是 {判断}，建议：
/{目标} — {一句话理由}
（其他可选：{备选1}、{备选2}）三行以内说完。用户要的是入口，不是综述。
/compare-masters 的 HARD-GATE 一致由目标 skill 自己的边界规则处理密法内容
| 传统 | 祖师 | |------|------| | 印度 | master-nagarjuna | | 汉传 | master-kumarajiva · master-zhiyi · master-fazang · master-xuanzang · master-huineng · master-yinguang · master-ouyi · master-xuyun | | 藏传 | master-atisha · master-tsongkhapa · master-milarepa | | 南传 | master-buddhaghosa · master-mahasi-sayadaw · master-ajahn-chah |
> 此表仅供快速定位。判断该选谁时以 routing.json 与各 meta.json 为准。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-21 | pass→pass | 20,494 | 15,693 | -23% | 1 | 1 | 0% | 2,879 | 3,175 | +10% | 0 | 0 | — |
case-01 | fail→fail | 10,240 | 3,740 | -63% | 1 | 1 | 0% | 1,278 | 1,728 | +35% | 0 | 0 | — |
case-02 | fail→fail | 9,984 | 4,291 | -57% | 1 | 1 | 0% | 1,766 | 1,899 | +8% | 0 | 0 | — |
case-03 | fail→fail | 8,111 | 6,865 | -15% | 1 | 1 | 0% | 1,403 | 2,245 | +60% | 0 | 0 | — |
case-04 | fail→pass | 9,189 | 3,275 | -64% | 1 | 1 | 0% | 1,551 | 1,549 | -0% | 0 | 0 | — |
case-05 | fail→fail | 23,498 | 3,930 | -83% | 1 | 1 | 0% | 3,626 | 1,799 | -50% | 0 | 0 | — |
case-06 | fail→fail | 24,443 | 4,500 | -82% | 1 | 1 | 0% | 3,604 | 1,748 | -51% | 0 | 0 | — |
case-07 | fail→fail | 25,907 | 4,135 | -84% | 1 | 1 | 0% | 2,869 | 1,745 | -39% | 0 | 0 | — |
case-08 | fail→fail | 17,919 | 4,640 | -74% | 1 | 1 | 0% | 2,608 | 1,807 | -31% | 0 | 0 | — |
case-09 | fail→fail | 23,324 | 4,473 | -81% | 1 | 1 | 0% | 3,576 | 1,735 | -51% | 0 | 0 | — |
case-10 | fail→fail | 22,785 | 5,865 | -74% | 1 | 1 | 0% | 3,345 | 1,916 | -43% | 0 | 0 | — |
case-11 | fail→pass | 22,738 | 4,339 | -81% | 1 | 1 | 0% | 3,270 | 1,709 | -48% | 0 | 0 | — |
case-12 | fail→fail | 20,208 | 5,315 | -74% | 1 | 1 | 0% | 2,970 | 1,825 | -39% | 0 | 0 | — |
case-13 | fail→pass | 23,087 | 4,509 | -80% | 1 | 1 | 0% | 3,375 | 1,758 | -48% | 0 | 0 | — |
case-14 | fail→fail | 23,764 | 4,398 | -81% | 1 | 1 | 0% | 3,709 | 1,750 | -53% | 0 | 0 | — |
case-15 | fail→pass | 22,981 | 3,618 | -84% | 1 | 1 | 0% | 3,472 | 1,542 | -56% | 0 | 0 | — |
case-16 | fail→fail | 20,992 | 4,157 | -80% | 1 | 1 | 0% | 2,912 | 1,695 | -42% | 0 | 0 | — |
case-17 | fail→pass | 20,293 | 7,603 | -63% | 1 | 1 | 0% | 2,799 | 2,247 | -20% | 0 | 0 | — |
case-18 | fail→fail | 19,647 | 7,894 | -60% | 1 | 1 | 0% | 2,717 | 2,292 | -16% | 0 | 0 | — |
case-19 | fail→pass | 18,666 | 7,110 | -62% | 1 | 1 | 0% | 2,816 | 2,148 | -24% | 0 | 0 | — |
case-20 | fail→fail | 26,142 | 4,019 | -85% | 1 | 1 | 0% | 3,376 | 1,748 | -48% | 0 | 0 | — |
case-22 | pass→fail | 32,490 | 9,513 | -71% | 1 | 1 | 0% | 4,836 | 2,703 | -44% | 0 | 0 | — |
case-23 | pass→fail | 34,681 | 9,545 | -72% | 1 | 1 | 0% | 5,012 | 2,587 | -48% | 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. 23 cases were attempted. The headline lift of +17 percentage points is the difference between those two pass rates over the 23 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/21/2026 | -26% |
Other measured skills in the registry, with their headline benchmark lift.