Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Clash Verge 诊断与配置管理。网络诊断、配置 AI 工具路由规则（Claude/Codex/Cursor/Warp）、查看/克隆订阅配置、切换配置文件。当用户遇到网络问题、要配置代理规则、切换订阅、查看 Clash 配置时使用
.claude/skills/majiayu000-clash-doctor/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 92% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -53% | 0% |
Inspect Clash Verge or mihomo, manage profile-specific routing, and diagnose connectivity.
references/full-guide.md for commands, templates, and diagnosis matrices.references/local-hub.md before hub, align, roles, clone, setup-ai, or switch, and whenever the user says a previous sync did not apply.tun.route-exclude-address is union-only. DIRECT is not a TUN bypass.PUT /configs does not re-merge overlays.| Argument | Mode | |---|---| | empty or a hostname | diagnose | | profiles / list / ls | list local and remote profiles | | status | live kernel, TUN, and selected nodes | | hub / roles | local-hub topology | | align / sync-check | read-only multi-machine comparison | | switch-check / health | pre-switch health | | clone / setup-ai / switch | confirmed profile writes | | rustdesk / bypass | TUN exclusion for a relay IP |
Resolve the named target before setup-ai. Apply airport-style enhancement overlays only when that target is type: remote; edit the named hub YAML when it is type: local.
chains[-1] is the matched group and chains[0] is the dialed node.references/full-guide.md — commands, templates, and diagnosis matrixreferences/local-hub.md — hub ownership, node roles, alignment, and landmines| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | pass→pass | 30,945 | 20,040 | -35% | 1 | 1 | 0% | 2,147 | 2,514 | +17% | 0 | 0 | — |
case-01 | fail→fail | 22,112 | 26,394 | +19% | 1 | 1 | 0% | 2,624 | 3,494 | +33% | 0 | 0 | — |
case-02 | fail→fail | 20,228 | 39,073 | +93% | 1 | 1 | 0% | 2,790 | 3,774 | +35% | 0 | 0 | — |
case-03 | fail→fail | 11,410 | 20,529 | +80% | 1 | 1 | 0% | 981 | 3,035 | +209% | 0 | 0 | — |
case-05 | pass→pass | 17,793 | 16,044 | -10% | 1 | 1 | 0% | 2,202 | 2,334 | +6% | 0 | 0 | — |
case-06 | pass→fail | 16,092 | 23,988 | +49% | 1 | 1 | 0% | 2,117 | 2,447 | +16% | 0 | 0 | — |
case-07 | pass→pass | 20,353 | 16,781 | -18% | 1 | 1 | 0% | 2,434 | 2,579 | +6% | 0 | 0 | — |
case-08 | pass→pass | 12,611 | 14,258 | +13% | 1 | 1 | 0% | 1,172 | 1,966 | +68% | 0 | 0 | — |
case-09 | fail→pass | 16,614 | 11,955 | -28% | 1 | 1 | 0% | 1,931 | 1,662 | -14% | 0 | 0 | — |
case-10 | pass→pass | 18,454 | 5,909 | -68% | 1 | 1 | 0% | 899 | 1,379 | +53% | 0 | 0 | — |
case-11 | fail→pass | 16,430 | 13,616 | -17% | 1 | 1 | 0% | 2,012 | 1,834 | -9% | 0 | 0 | — |
case-12 | pass→pass | 16,247 | 7,785 | -52% | 1 | 1 | 0% | 1,658 | 1,490 | -10% | 0 | 0 | — |
case-13 | fail→pass | 18,343 | 11,096 | -40% | 1 | 1 | 0% | 2,119 | 1,578 | -26% | 0 | 0 | — |
case-14 | fail→pass | 27,015 | 24,764 | -8% | 1 | 1 | 0% | 1,563 | 3,004 | +92% | 0 | 0 | — |
case-15 | fail→pass | 28,067 | 7,717 | -73% | 1 | 1 | 0% | 2,112 | 999 | -53% | 0 | 0 | — |
case-16 | fail→pass | 17,472 | 7,533 | -57% | 1 | 1 | 0% | 2,012 | 946 | -53% | 0 | 0 | — |
case-17 | pass→pass | 16,594 | 12,885 | -22% | 1 | 1 | 0% | 1,756 | 1,056 | -40% | 0 | 0 | — |
case-18 | pass→pass | 22,567 | 13,090 | -42% | 1 | 1 | 0% | 2,688 | 1,384 | -49% | 0 | 0 | — |
case-19 | fail→pass | 17,821 | 8,706 | -51% | 1 | 1 | 0% | 1,865 | 1,015 | -46% | 0 | 0 | — |
case-20 | fail→pass | 19,416 | 10,768 | -45% | 1 | 1 | 0% | 1,638 | 1,424 | -13% | 0 | 0 | — |
case-21 | pass→pass | 7,885 | 16,747 | +112% | 1 | 1 | 0% | 1,378 | 1,181 | -14% | 0 | 0 | — |
case-22 | fail→pass | 17,867 | 11,512 | -36% | 1 | 1 | 0% | 2,063 | 2,310 | +12% | 0 | 0 | — |
case-23 | fail→pass | 25,560 | 12,114 | -53% | 1 | 1 | 0% | 2,374 | 1,619 | -32% | 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 +39 percentage points is the difference between those two pass rates over the 23 comparable cases. 1 case got worse with the skill loaded, and it is 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 | 8/24/2026 | — |
Other measured skills in the registry, with their headline benchmark lift.