Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →基于退哥短线交易规则的A股场景化决策技能。Use when 用户要按交易场景查看短线规则、做选股、判断趋势回踩、涨停回调、连板接力、洗盘结束、卖出失效或仓位纪律。
.claude/skills/shouldnotappearcalm-tuige-shortline-trading/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 71% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 19% | 0% |
这个 skill 用来把退哥体系中的短线规则整理成可复用的决策框架，重点回答 4 类问题：
glossary.mdapi-reference/market-regime.mdapi-reference/stock-selection.mdapi-reference/trend-setups.mdapi-reference/limit-up-pullback-setups.mdapi-reference/relay-setups.mdapi-reference/washout-breakout-setups.mdapi-reference/exit-failure-rules.mdapi-reference/position-discipline.md输出时统一包含以下结构：
单只标的建议使用以下字段组织：
当前目录已提供可执行脚本，采用统一的 daily_decisions / realtime_quotes / strategy_lab 结构：
scripts/daily_decisions.pytrend_pullback 信号from_previous_day_close 与 from_last_close8（可通过 --max-holdings 下调）scripts/realtime_quotes.pyscripts/strategy_lab/strategy_params.pyscripts/strategy_lab/strategies.pybashSKILL_DIR="<本 skill 绝对路径>" python3 "$SKILL_DIR/scripts/daily_decisions.py" --json
常用参数：
--top-n：股票池大小，默认 120--max-holdings：最大持仓上限，默认 8--max-buys：买入输出上限，默认 8（最终会再受 max-holdings 约束）--holdings：持仓文件路径，一行一个代码--roundtrip-cost-bps：成本过滤，默认 45--entry-consensus-min：鲁棒一致性阈值，默认 0.67--disable-robust-check：关闭鲁棒性检查（仅调试）tuige_shortline_v1，优先落地“可量化且可复测”的部分。| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 28,379 | 27,754 | -2% | 1 | 1 | 0% | 4,105 | 5,597 | +36% | 0 | 0 | — |
case-02 | fail→pass | 21,792 | 27,351 | +26% | 1 | 1 | 0% | 3,361 | 5,744 | +71% | 0 | 0 | — |
case-03 | fail→fail | 19,690 | 31,875 | +62% | 1 | 1 | 0% | 2,994 | 5,726 | +91% | 0 | 0 | — |
case-04 | pass→pass | 16,761 | 10,572 | -37% | 1 | 1 | 0% | 2,786 | 2,556 | -8% | 0 | 0 | — |
case-05 | fail→fail | 27,175 | 47,284 | +74% | 1 | 1 | 0% | 5,603 | 9,185 | +64% | 0 | 0 | — |
case-06 | pass→pass | 14,755 | 15,980 | +8% | 1 | 1 | 0% | 2,022 | 3,591 | +78% | 0 | 0 | — |
case-07 | fail→pass | 13,862 | 2,732 | -80% | 1 | 1 | 0% | 1,862 | 1,667 | -10% | 0 | 0 | — |
case-08 | fail→pass | 13,250 | 3,185 | -76% | 1 | 1 | 0% | 2,118 | 1,813 | -14% | 0 | 0 | — |
case-09 | fail→pass | 14,200 | 2,893 | -80% | 1 | 1 | 0% | 2,463 | 1,707 | -31% | 0 | 0 | — |
case-10 | fail→pass | 11,548 | 3,643 | -68% | 1 | 1 | 0% | 1,545 | 1,838 | +19% | 0 | 0 | — |
case-11 | fail→pass | 13,198 | 3,998 | -70% | 1 | 1 | 0% | 2,086 | 1,916 | -8% | 0 | 0 | — |
case-12 | pass→pass | 17,171 | 14,132 | -18% | 1 | 1 | 0% | 2,405 | 3,302 | +37% | 0 | 0 | — |
case-13 | pass→pass | 24,177 | 19,576 | -19% | 1 | 1 | 0% | 3,001 | 4,121 | +37% | 0 | 0 | — |
case-14 | pass→pass | 16,209 | 18,737 | +16% | 1 | 1 | 0% | 2,427 | 4,137 | +70% | 0 | 0 | — |
case-15 | pass→pass | 20,961 | 22,763 | +9% | 1 | 1 | 0% | 3,039 | 4,857 | +60% | 0 | 0 | — |
case-16 | fail→pass | 19,466 | 15,091 | -22% | 1 | 1 | 0% | 2,646 | 3,469 | +31% | 0 | 0 | — |
case-17 | pass→pass | 15,240 | 15,455 | +1% | 1 | 1 | 0% | 2,205 | 3,443 | +56% | 0 | 0 | — |
case-18 | pass→pass | 22,364 | 17,337 | -22% | 1 | 1 | 0% | 2,657 | 3,950 | +49% | 0 | 0 | — |
case-19 | pass→pass | 14,479 | 6,743 | -53% | 1 | 1 | 0% | 1,891 | 2,298 | +22% | 0 | 0 | — |
case-20 | fail→pass | 15,897 | 4,225 | -73% | 1 | 1 | 0% | 2,244 | 1,850 | -18% | 0 | 0 | — |
case-21 | fail→pass | 13,631 | 4,231 | -69% | 1 | 1 | 0% | 2,098 | 1,868 | -11% | 0 | 0 | — |
case-22 | fail→pass | 10,675 | 2,344 | -78% | 1 | 1 | 0% | 1,433 | 1,661 | +16% | 0 | 0 | — |
case-23 | pass→pass | 10,179 | 2,763 | -73% | 1 | 1 | 0% | 1,576 | 1,810 | +15% | 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 +43 percentage points is the difference between those two pass rates over the 23 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.