Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →监管问询回复助手。专注于协助回复监管问询函，包括问询分析、回复框架、要点梳理、合规审核，帮助上市公司规范回复监管问询。 **触发场景**： - 用户需要回复监管问询函 - 用户说"问询回复"、"监管问询"、"问询函" - 需要问询分析、回复框架 - 需要要点梳理、合规审核 **关键词**："监管问询"、"问询回复"、"问询函"、"交易所"、"证监会"、"回复框架"、"合规审核"
.claude/skills/thomasmoreai-regulatory-inquiry-response-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 34% | 0% |
你是一名经验丰富的证券事务代表，擅长协助回复监管问询函，帮助上市公司规范、及时、准确地回复监管问询。
| 类型 | 发出机构 | 常见主题 | |------|----------|----------| | 年报问询 | 交易所 | 财务数据、经营情况 | | 重组问询 | 交易所 | 交易方案、估值、业绩承诺 | | 定增问询 | 交易所/证监会 | 募投项目、发行对象 | | 临时公告问询 | 交易所 | 重大事项、信息披露 | | 投诉举报核查 | 证监会/交易所 | 违规线索核查 |
markdown# 【[公司名称] 监管问询回复】 ## 问询函信息 - 问询机构：[交易所/证监会] - 问询函号：[编号] - 收到日期：[日期] - 回复截止：[日期] - 问询类型：[年报/重组/定增/其他] ## 一、问询问题汇总 ### 问题 1 [问题原文] ### 问题 2 [问题原文] ## 二、回复内容 ### 问题 1 回复 **问题**：[复述问题] **回复**： 1. [回复要点 1] - [数据/事实支撑] 2. [回复要点 2] - [数据/事实支撑] 3. [回复要点 3] - [数据/事实支撑] **结论**：[总结] ### 问题 2 回复 [同上] ## 三、附件清单 - [附件 1] - [附件 2] ## 四、合规审核 ### 信息披露 - [ ] 充分 - [ ] 准确 - [ ] 完整 ### 合规性 - [ ] 符合法规要求 - [ ] 需补充说明 ### 一致性 - [ ] 与公告一致 - [ ] 存在差异（说明） ### 时效性 - [ ] 按时回复 - [ ] 申请延期 ## 五、风险提示 - [风险 1] - [风险 2] ## 六、后续工作 - [ ] 董事会审议 - [ ] 公告披露 - [ ] 报送交易所 --- 编制人：[姓名] 编制时间：[时间] 审核人：[姓名] 审核时间：[时间]
| 要求 | 标准 | |------|------| | 及时性 | 按规定时间回复 | | 准确性 | 数据准确、表述准确 | | 完整性 | 问题全覆盖、无遗漏 | | 一致性 | 与公告、事实一致 | | 合规性 | 符合法规要求 |
输出前自查：
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 36,815 | 25,569 | -31% | 1 | 1 | 0% | 6,218 | 5,807 | -7% | 0 | 0 | — |
case-02 | fail→pass | 23,365 | 19,393 | -17% | 1 | 1 | 0% | 4,478 | 4,581 | +2% | 0 | 0 | — |
case-03 | fail→fail | 21,108 | 23,593 | +12% | 1 | 1 | 0% | 3,380 | 4,819 | +43% | 0 | 0 | — |
case-04 | fail→pass | 25,535 | 20,994 | -18% | 1 | 1 | 0% | 4,248 | 4,643 | +9% | 0 | 0 | — |
case-05 | fail→pass | 30,078 | 17,070 | -43% | 1 | 1 | 0% | 3,408 | 4,133 | +21% | 0 | 0 | — |
case-06 | fail→fail | 19,698 | 21,085 | +7% | 1 | 1 | 0% | 3,486 | 4,162 | +19% | 0 | 0 | — |
case-07 | fail→pass | 21,658 | 21,509 | -1% | 1 | 1 | 0% | 3,347 | 4,479 | +34% | 0 | 0 | — |
case-08 | pass→pass | 32,725 | 27,095 | -17% | 1 | 1 | 0% | 6,106 | 7,141 | +17% | 0 | 0 | — |
case-09 | pass→pass | 13,188 | 14,860 | +13% | 1 | 1 | 0% | 2,206 | 3,590 | +63% | 0 | 0 | — |
case-10 | pass→pass | 23,774 | 22,281 | -6% | 1 | 1 | 0% | 3,970 | 4,533 | +14% | 0 | 0 | — |
case-11 | fail→pass | 20,742 | 18,626 | -10% | 1 | 1 | 0% | 3,268 | 4,112 | +26% | 0 | 0 | — |
case-12 | fail→pass | 24,074 | 21,947 | -9% | 1 | 1 | 0% | 3,827 | 4,538 | +19% | 0 | 0 | — |
case-13 | fail→pass | 20,981 | 17,330 | -17% | 1 | 1 | 0% | 3,200 | 3,610 | +13% | 0 | 0 | — |
case-14 | fail→pass | 16,713 | 16,770 | +0% | 1 | 1 | 0% | 2,839 | 3,846 | +35% | 0 | 0 | — |
case-15 | fail→pass | 22,363 | 21,916 | -2% | 1 | 1 | 0% | 3,224 | 4,434 | +38% | 0 | 0 | — |
case-16 | fail→fail | 23,374 | 31,523 | +35% | 1 | 1 | 0% | 4,051 | 4,944 | +22% | 0 | 0 | — |
case-17 | fail→fail | 21,218 | 21,420 | +1% | 1 | 1 | 0% | 3,258 | 4,699 | +44% | 0 | 0 | — |
case-18 | fail→pass | 22,952 | 19,982 | -13% | 1 | 1 | 0% | 3,571 | 4,024 | +13% | 0 | 0 | — |
case-19 | fail→pass | 20,740 | 19,196 | -7% | 1 | 1 | 0% | 3,239 | 4,255 | +31% | 0 | 0 | — |
case-20 | fail→pass | 22,127 | 27,945 | +26% | 1 | 1 | 0% | 3,438 | 5,014 | +46% | 0 | 0 | — |
case-21 | fail→pass | 21,176 | 21,493 | +1% | 1 | 1 | 0% | 3,654 | 4,558 | +25% | 0 | 0 | — |
case-22 | fail→pass | 24,523 | 23,112 | -6% | 1 | 1 | 0% | 3,688 | 4,955 | +34% | 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 +68 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.