Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →基于“均线定方向，MACD定节奏”的A股选股与交易计划技能。Use when 用户要求把均线与MACD结合做选股、筛票、盘前候选池、趋势跟随、回踩再上、金叉确认、顶背离减仓、或希望把技术判断沉淀成可执行规则与风控模板。
.claude/skills/shouldnotappearcalm-macd-trend-resonance-stock-picker/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 88% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 89% | 0% |
| case-04 | ✓→✗ | ▼ Worse | -49% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 72% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 31% | 0% |
把“均线看趋势，MACD看节奏”落成一套可执行的选股流程，输出：
../a-share-skill/scripts/ 获取，禁止换未验证数据源。fetch_realtime.py 获取实时快照fetch_history.py 获取日线/60分钟K线fetch_technical.py 获取 MACD、均线等技术指标fetch_stock_events.py 获取重大事件与异动背景先明确：
日线定方向 + 60分钟找触发若用户没给范围：
核心规则：
60日线向上 + 收盘价在60日线上方 → 趋势合格60日线走平或粘合 → 仅观察，不列强候选60日线向下 + 收盘价在60日线下方 → 直接剔除可补充观察：
方向是硬过滤条件。方向不对，后续 MACD 再漂亮也不能进强候选池。
优先级从高到低：
MACD 位于 0 轴上方金叉红柱缩短后重新放大DIF/DEA 同步向上60分钟 MACD 与日线同向降级或剔除条件：
同时满足：
动作：
满足趋势条件，同时：
动作：
如：
动作：
任一命中：
动作：
默认只给“条件触发式”买点，不凭主观感觉提前下手。
优先触发：
每次都必须给：
常用失效定义：
优先减仓信号：
直接离场信号：
个股结论前，至少补看：
若板块与指数明显逆风：
动作映射：
按以下结构输出：
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 24,530 | 83,494 | +240% | 1 | 1 | 0% | 4,066 | 10,649 | +162% | 0 | 0 | — |
case-02 | fail→fail | 29,324 | 73,122 | +149% | 1 | 1 | 0% | 4,175 | 2,891 | -31% | 0 | 0 | — |
case-03 | fail→fail | 25,710 | 8,949 | -65% | 1 | 1 | 0% | 4,146 | 2,442 | -41% | 0 | 0 | — |
case-04 | pass→fail | 23,626 | 8,505 | -64% | 1 | 1 | 0% | 4,467 | 2,290 | -49% | 0 | 0 | — |
case-05 | pass→pass | 13,939 | 17,319 | +24% | 1 | 1 | 0% | 2,513 | 4,315 | +72% | 0 | 0 | — |
case-06 | pass→pass | 19,128 | 23,716 | +24% | 1 | 1 | 0% | 4,245 | 5,540 | +31% | 0 | 0 | — |
case-07 | fail→pass | 18,946 | 14,955 | -21% | 1 | 1 | 0% | 2,294 | 4,317 | +88% | 0 | 0 | — |
case-08 | pass→pass | 19,781 | 28,563 | +44% | 1 | 1 | 0% | 2,741 | 5,737 | +109% | 0 | 0 | — |
case-09 | pass→pass | 21,295 | 18,818 | -12% | 1 | 1 | 0% | 2,407 | 4,270 | +77% | 0 | 0 | — |
case-10 | pass→pass | 15,504 | 11,875 | -23% | 1 | 1 | 0% | 2,056 | 3,569 | +74% | 0 | 0 | — |
case-11 | fail→pass | 13,841 | 16,111 | +16% | 1 | 1 | 0% | 2,160 | 4,088 | +89% | 0 | 0 | — |
case-12 | pass→pass | 7,707 | 2,737 | -64% | 1 | 1 | 0% | 1,235 | 2,221 | +80% | 0 | 0 | — |
case-13 | pass→pass | 18,861 | 23,225 | +23% | 1 | 1 | 0% | 2,524 | 4,971 | +97% | 0 | 0 | — |
case-14 | pass→pass | 19,075 | 17,191 | -10% | 1 | 1 | 0% | 2,491 | 4,036 | +62% | 0 | 0 | — |
case-15 | pass→pass | 20,684 | 18,102 | -12% | 1 | 1 | 0% | 2,528 | 4,250 | +68% | 0 | 0 | — |
case-16 | pass→pass | 21,750 | 10,292 | -53% | 1 | 1 | 0% | 2,668 | 3,402 | +28% | 0 | 0 | — |
case-17 | pass→pass | 16,277 | 15,410 | -5% | 1 | 1 | 0% | 2,187 | 4,122 | +88% | 0 | 0 | — |
case-18 | pass→pass | 24,538 | 17,487 | -29% | 1 | 1 | 0% | 3,373 | 4,331 | +28% | 0 | 0 | — |
case-19 | pass→pass | 16,475 | 14,485 | -12% | 1 | 1 | 0% | 2,186 | 3,477 | +59% | 0 | 0 | — |
case-20 | pass→pass | 21,413 | 7,994 | -63% | 1 | 1 | 0% | 3,058 | 3,010 | -2% | 0 | 0 | — |
case-21 | pass→pass | 18,825 | 15,399 | -18% | 1 | 1 | 0% | 2,420 | 3,818 | +58% | 0 | 0 | — |
case-22 | pass→pass | 15,347 | 22,914 | +49% | 1 | 1 | 0% | 1,860 | 2,196 | +18% | 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, and 18 counted toward the lift figure. The other 4 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 +5 percentage points is the difference between those two pass rates over the 18 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.