Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股监管追踪/处罚公告/问询函。当用户说"监管"、"问询函"、"处罚"、"立案调查"、"证监会"、"交易所"、"关注函"、"警示函"、"regulatory"、"XX被监管了吗"、"XX收到问询函"、"违规"时触发。追踪上市公司监管动态（问询函/关注函/警示函/行政处罚/立案调查），分析监管风险对股价和基本面的影响。支持研报风格（formal）和快速预警风格（brief）。
.claude/skills/aifinlab-a-share-regulatory-watch/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-12 | ✓→✗ | ▼ Worse | -33% | 0% |
ak.stock_notice_report(symbol='CODE') 获取上市公司公告（如可用）"{公司名}" 问询函 / 关注函 / 警示函"{公司名}" 行政处罚 / 立案调查"{公司名}" 证监会 / 交易所 / 证监局"{公司名}" 违规 / 信息披露对照 references/regulatory-guide.md 中的监管措施等级，对检索到的监管事件进行分级：
| 等级 | 措施类型 | 严重性 | 标记 | |------|---------|--------|------| | L1 | 问询函 | 低 | 📋 | | L2 | 关注函 | 中低 | 📌 | | L3 | 警示函 / 监管谈话 | 中 | ⚠️ | | L4 | 行政监管措施（限制业务/责令改正） | 中高 | 🔶 | | L5 | 立案调查 | 高 | 🔴 | | L6 | 行政处罚（罚款/市场禁入） | 极高 | ⛔ |
标注每个事件的监管层级（证监会 > 交易所 > 地方证监局）
将监管事项归类并分析实质：
重点关注：
股价影响分析：
基本面影响判断：
退市风险评估（参考退市新规）：
brief 模式（默认）：
formal 模式：
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 22,209 | 10,890 | -51% | 1 | 1 | 0% | 2,829 | 1,877 | -34% | 0 | 0 | — |
case-02 | fail→fail | 33,773 | 10,538 | -69% | 1 | 1 | 0% | 4,922 | 1,947 | -60% | 0 | 0 | — |
case-03 | fail→fail | 36,625 | 10,667 | -71% | 1 | 1 | 0% | 3,263 | 1,871 | -43% | 0 | 0 | — |
case-04 | fail→fail | 21,369 | 7,075 | -67% | 1 | 1 | 0% | 2,969 | 1,739 | -41% | 0 | 0 | — |
case-05 | fail→fail | 25,001 | 9,669 | -61% | 1 | 1 | 0% | 3,674 | 1,835 | -50% | 0 | 0 | — |
case-06 | fail→pass | 18,995 | 14,732 | -22% | 1 | 1 | 0% | 2,526 | 3,604 | +43% | 0 | 0 | — |
case-07 | fail→fail | 15,915 | 18,040 | +13% | 1 | 1 | 0% | 2,405 | 2,917 | +21% | 0 | 0 | — |
case-08 | fail→fail | 17,725 | 17,713 | -0% | 1 | 1 | 0% | 2,736 | 4,059 | +48% | 0 | 0 | — |
case-09 | pass→pass | 18,481 | 8,402 | -55% | 1 | 1 | 0% | 1,832 | 2,714 | +48% | 0 | 0 | — |
case-16 | fail→pass | 63,334 | 5,073 | -92% | 1 | 1 | 0% | 4,262 | 2,197 | -48% | 0 | 0 | — |
case-10 | pass→pass | 17,243 | 15,767 | -9% | 1 | 1 | 0% | 2,644 | 3,836 | +45% | 0 | 0 | — |
case-11 | pass→pass | 15,227 | 17,639 | +16% | 1 | 1 | 0% | 2,256 | 4,098 | +82% | 0 | 0 | — |
case-12 | pass→fail | 18,814 | 8,773 | -53% | 1 | 1 | 0% | 2,663 | 1,777 | -33% | 0 | 0 | — |
case-13 | pass→pass | 21,166 | 19,304 | -9% | 1 | 1 | 0% | 2,559 | 4,009 | +57% | 0 | 0 | — |
case-14 | fail→pass | 21,650 | 17,612 | -19% | 1 | 1 | 0% | 3,151 | 4,007 | +27% | 0 | 0 | — |
case-15 | fail→fail | 11,081 | 8,781 | -21% | 1 | 1 | 0% | 1,540 | 1,652 | +7% | 0 | 0 | — |
case-17 | pass→pass | 18,915 | 22,895 | +21% | 1 | 1 | 0% | 3,081 | 4,510 | +46% | 0 | 0 | — |
case-18 | fail→pass | 16,885 | 20,899 | +24% | 1 | 1 | 0% | 2,351 | 2,780 | +18% | 0 | 0 | — |
case-19 | pass→pass | 13,718 | 15,737 | +15% | 1 | 1 | 0% | 2,306 | 3,820 | +66% | 0 | 0 | — |
case-20 | pass→pass | 26,349 | 27,143 | +3% | 1 | 1 | 0% | 3,334 | 5,303 | +59% | 0 | 0 | — |
case-21 | pass→pass | 21,599 | 21,862 | +1% | 1 | 1 | 0% | 3,269 | 4,869 | +49% | 0 | 0 | — |
case-22 | pass→fail | 23,121 | 15,202 | -34% | 1 | 1 | 0% | 3,490 | 2,204 | -37% | 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 14 counted toward the lift figure. The other 8 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 +9 percentage points is the difference between those two pass rates over the 14 comparable cases. 6 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.