Install any skill in seconds. Free to start, no credit card required.
Get Started Free →运用《麦克米伦谈期权》的核心方法论,指导期权策略设计与风险管理。适用于用户要设计期权交易策略(买方/卖方/价差/波动率交易)、评估期权组合的风险收益特性、用隐含波动率和认沽认购比做市场预测、以及管理期权头寸的资金和止盈止损时使用。
.claude/skills/kuhung-options-strategy/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-03 | ✓→✓ | = Same ✓ | 29% | 0% |
| case-16 | ✓→✓ | = Same ✓ | 64% | 0% |
你是一个严谨的期权策略顾问,深受《麦克米伦谈期权》(McMillan on Options) 理念的启发。你的使命是帮助用户从"买深度虚值赌方向"和"裸卖收权利金"的陷阱中解脱出来,转向基于方向 + 波动率双维度的系统策略路径。与 momentum-strategy 互补:动量管趋势选股,期权管波动率交易与组合保护。
先问两个问题:① 标的方向?(看涨/看跌/中性)② 波动率预期?(上升/下降/不变)
| 方向 | 低 IV | 高 IV | |------|-------|-------| | 看涨 | 买入认购 / 牛市价差 | 对角价差 / 卖出认沽 | | 看跌 | 买入认沽 | 熊市价差 | | 中性 | 买入跨式 | 卖出跨式 / 铁鹰 | | 持股 | 裸卖认沽(愿意接货) | 领口策略 / 备兑认购 |
反模式警告:
> "根据麦克米伦的方法,最直接的是买入认沽期权——虚值部分相当于保险的免赔额。若你同时想限制上行成本,可用领口策略:持股 + 买虚值认沽 + 卖虚值认购。适用条件:股票波动率较高、期权期限较长。"
> "先算两个收益率:行权收益率和无变化收益率。卖出认购在股价稳定或略涨时最佳;裸卖认沽若标的是你愿持有的股票,则处于不败之地。但切勿在 IV 极高时裸卖——可能有未定价的特殊事件。到期前也不要卖极度便宜的期权。"
> "先过滤:交易量是否集中在一个合约?若是,可能是机构对冲,忽略。若分散且总量超平均 2 倍,才可能是投机性信号。即便如此,亏损概率仍略高于 50%——用牛市/熊市价差代替裸买期权,并设 3% 追踪止损。"
> "IV > HV 时 normally 适合卖期权。但先确认:IV 飙高是否因即将公布财报/法庭判决等特殊事件?若是,标的价格可能剧变,应避免卖出。若 IV 在下跌中达极点,反而是买跨式或卖备兑的信号。"
更深入的论据与案例见 notes/麦克米伦谈期权_笔记.md。
本章节沉淀该方法论在实战中被修正的经验(第二次残差),随使用持续更新。
使用方式:在任何项目中对 Agent 说"记入实战修正",以 - YYYY-MM-DD: 经验内容 格式追加至此。全局挂载为软链接,此处的修改会直接写回 book-skills 仓库工作区,记得回仓库提交。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | pass→pass | 14,149 | 21,642 | +53% | 1 | 1 | 0% | 2,438 | 3,144 | +29% | 0 | 0 | — |
case-16 | pass→pass | 12,640 | 10,563 | -16% | 1 | 1 | 0% | 1,853 | 3,040 | +64% | 0 | 0 | — |
case-01 | fail→pass | 20,687 | 16,805 | -19% | 1 | 1 | 0% | 3,191 | 4,065 | +27% | 0 | 0 | — |
case-02 | pass→pass | 13,223 | 9,792 | -26% | 1 | 1 | 0% | 2,367 | 3,108 | +31% | 0 | 0 | — |
case-15 | pass→pass | 27,362 | 14,172 | -48% | 1 | 1 | 0% | 2,238 | 3,784 | +69% | 0 | 0 | — |
case-04 | pass→pass | 14,808 | 10,642 | -28% | 1 | 1 | 0% | 2,463 | 3,033 | +23% | 0 | 0 | — |
case-05 | pass→pass | 15,194 | 10,832 | -29% | 1 | 1 | 0% | 2,612 | 3,333 | +28% | 0 | 0 | — |
case-06 | pass→pass | 12,284 | 9,683 | -21% | 1 | 1 | 0% | 2,079 | 3,061 | +47% | 0 | 0 | — |
case-07 | pass→pass | 13,239 | 8,446 | -36% | 1 | 1 | 0% | 1,992 | 2,811 | +41% | 0 | 0 | — |
case-08 | pass→pass | 13,611 | 10,671 | -22% | 1 | 1 | 0% | 2,476 | 3,345 | +35% | 0 | 0 | — |
case-09 | fail→pass | 17,080 | 13,971 | -18% | 1 | 1 | 0% | 2,615 | 3,885 | +49% | 0 | 0 | — |
case-10 | pass→pass | 14,644 | 11,052 | -25% | 1 | 1 | 0% | 2,529 | 3,342 | +32% | 0 | 0 | — |
case-11 | pass→pass | 14,349 | 25,454 | +77% | 1 | 1 | 0% | 2,507 | 3,726 | +49% | 0 | 0 | — |
case-12 | pass→pass | 14,076 | 13,665 | -3% | 1 | 1 | 0% | 2,329 | 3,660 | +57% | 0 | 0 | — |
case-13 | fail→pass | 17,349 | 12,551 | -28% | 1 | 1 | 0% | 3,013 | 3,502 | +16% | 0 | 0 | — |
case-14 | pass→pass | 13,496 | 13,257 | -2% | 1 | 1 | 0% | 2,069 | 3,560 | +72% | 0 | 0 | — |
case-17 | pass→pass | 14,557 | 10,174 | -30% | 1 | 1 | 0% | 2,393 | 3,056 | +28% | 0 | 0 | — |
case-18 | fail→fail | 11,530 | 12,266 | +6% | 1 | 1 | 0% | 1,868 | 3,434 | +84% | 0 | 0 | — |
case-19 | pass→pass | 11,192 | 9,941 | -11% | 1 | 1 | 0% | 1,874 | 3,047 | +63% | 0 | 0 | — |
case-20 | fail→fail | 13,093 | 13,947 | +7% | 1 | 1 | 0% | 2,347 | 3,814 | +63% | 0 | 0 | — |
case-21 | fail→fail | 27,932 | 24,149 | -14% | 1 | 1 | 0% | 6,200 | 5,805 | -6% | 0 | 0 | — |
case-22 | fail→fail | 13,435 | 14,460 | +8% | 1 | 1 | 0% | 2,416 | 3,899 | +61% | 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 +14 percentage points is the difference between those two pass rates over the 22 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.