Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create Chinese monospaced compass-style PDCA or CAPD problem-solving cards using East-West-South-North-Center layout, full-width and half-width spaces, diagonal Unicode arrows, and strict Chinese character limits. Use when Codex needs to turn a problem, incident, decision tradeoff, root-cause analysis, improvement cycle, or CAPD/PDCA workflow into a square text diagram, Hermes/Discord-ready card, or Chinese 方位九宮圖 with center ◎, north Plan, east Do, south Check, west Action, and four diagonal tra
.claude/skills/twhsi-pdca/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 252% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 202% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 223% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 15% | 0% |
Use this skill to transform a problem into a Chinese text diagram shaped like a compass: North, East, South, West, four diagonal arrows, and a center ◎ core. The output should feel like a fixed-width visual card, not a prose explanation.
text code block for the final diagram. for large horizontal alignment and half-width spaces only for small adjustments.◎.北: 【P|Plan 計畫】 for goals, options, choices, hypotheses, and intended destination.東: 【D|Do 執行】 for concrete measures, steps,现场 action, and use of methods.南: 【C|Check 檢查】 for facts, results, confirmation, and next-cycle evidence.西: 【A|Action 改善】 for cause discovery, root-cause pursuit, correction, and institutional repair.中: 【◎核心】 for the problem essence, incident, contradiction, or decision point.For CAPD, the cycle begins at Check:
Check -> Action -> Plan -> Do -> CheckFor PDCA, the cycle begins at Plan:
Plan -> Do -> Check -> Action -> PlanUse 45-degree Unicode arrows at the four corners:
↗ means Action -> Plan: root causes become better plans.↘ means Plan -> Do: plans become concrete execution.↙ means Do -> Check: execution produces results to inspect.↖ means Check -> Action: inspection returns to root-cause improvement.Place arrows between the compass quadrants with enough full-width spacing to preserve the visual loop.
Adapt this template. Keep line lengths visually balanced rather than mathematically perfect.
text北 【P|Plan 計畫】 方案一:____ 方案二:____ 選 擇:____ ↗ ↘ 西【A|Action】 中【◎核心】 東【D|Do】 原因發現 ◎____ 手段活用 ──────── ──────── ──────── ____ ____ ____ ____ ____ ____ ____ ____ ____ ↖ ↙ 【C|Check 檢查】 事 實:____ 結 果:____ 下一輪:____ 南
text code block with full-width spacing and arrows.When content is too long, compress meaning before widening the diagram:
老虎已經逃出籠子 -> 老虎逃走人命安全與動物保護之間的取捨 -> 人獸取捨封鎖現場並避免民眾靠近 -> 封鎖現場防止同類事件再次發生 -> 防止再發Prefer four-character and six-character Chinese compounds when possible. Use eight characters only when needed.
text北 【P|Plan 計畫】 方案A:人命尊重 方案B:動物愛護 選 擇:先保人命 ↗ ↘ 西【A|Action】 中【◎核心】 東【D|Do】 原因發現 ◎老虎逃走 手段活用 ──────── ──────── ──────── 管理失誤 人獸危險 封鎖現場 機制失效 取捨判斷 搜尋活捉 根因追究 安全優先 必要處置 ↖ ↙ 【C|Check 檢查】 事 實:老虎逃出 結 果:安全控制 下一輪:防止再發 南
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 30,423 | 16,536 | -46% | 1 | 1 | 0% | 4,890 | 6,164 | +26% | 0 | 0 | — |
case-02 | fail→pass | 21,602 | 14,490 | -33% | 1 | 1 | 0% | 3,915 | 5,830 | +49% | 0 | 0 | — |
case-03 | fail→fail | 27,696 | 14,975 | -46% | 1 | 1 | 0% | 4,957 | 5,887 | +19% | 0 | 0 | — |
case-04 | pass→fail | 16,190 | 19,597 | +21% | 1 | 1 | 0% | 2,551 | 6,483 | +154% | 0 | 0 | — |
case-05 | pass→fail | 13,071 | 23,190 | +77% | 1 | 1 | 0% | 2,063 | 6,999 | +239% | 0 | 0 | — |
case-06 | pass→fail | 9,739 | 17,005 | +75% | 1 | 1 | 0% | 1,594 | 6,102 | +283% | 0 | 0 | — |
case-07 | pass→pass | 16,323 | 14,788 | -9% | 1 | 1 | 0% | 1,869 | 5,666 | +203% | 0 | 0 | — |
case-08 | fail→pass | 11,700 | 15,351 | +31% | 1 | 1 | 0% | 1,640 | 5,776 | +252% | 0 | 0 | — |
case-09 | pass→pass | 35,852 | 2,655 | -93% | 1 | 1 | 0% | 6,152 | 3,101 | -50% | 0 | 0 | — |
case-10 | fail→pass | 6,558 | 2,300 | -65% | 1 | 1 | 0% | 996 | 3,006 | +202% | 0 | 0 | — |
case-11 | fail→pass | 5,815 | 3,434 | -41% | 1 | 1 | 0% | 987 | 3,184 | +223% | 0 | 0 | — |
case-12 | fail→pass | 36,156 | 21,732 | -40% | 1 | 1 | 0% | 6,168 | 7,122 | +15% | 0 | 0 | — |
case-13 | fail→pass | 9,524 | 4,485 | -53% | 1 | 1 | 0% | 1,467 | 3,392 | +131% | 0 | 0 | — |
case-14 | fail→pass | 8,910 | 5,312 | -40% | 1 | 1 | 0% | 1,358 | 3,451 | +154% | 0 | 0 | — |
case-15 | fail→pass | 9,482 | 6,374 | -33% | 1 | 1 | 0% | 1,437 | 3,754 | +161% | 0 | 0 | — |
case-16 | fail→pass | 9,542 | 6,288 | -34% | 1 | 1 | 0% | 1,394 | 3,551 | +155% | 0 | 0 | — |
case-17 | fail→pass | 12,391 | 13,374 | +8% | 1 | 1 | 0% | 2,296 | 5,580 | +143% | 0 | 0 | — |
case-18 | fail→fail | 15,499 | 1,709 | -89% | 1 | 1 | 0% | 2,329 | 2,897 | +24% | 0 | 0 | — |
case-19 | fail→pass | 32,411 | 14,431 | -55% | 1 | 1 | 0% | 6,168 | 5,560 | -10% | 0 | 0 | — |
case-20 | fail→pass | 35,966 | 14,170 | -61% | 1 | 1 | 0% | 6,165 | 5,666 | -8% | 0 | 0 | — |
case-21 | fail→pass | 11,970 | 13,579 | +13% | 1 | 1 | 0% | 1,722 | 5,518 | +220% | 0 | 0 | — |
case-22 | pass→pass | 7,292 | 2,741 | -62% | 1 | 1 | 0% | 1,227 | 3,048 | +148% | 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. The headline lift of +45 percentage points is the difference between those two pass rates over the 22 comparable cases. 3 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.