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（再融资版）- 协助撰写再融资审核问询回复。当用户需要为定增、配股、可转债等再融资项目准备交易所/证监会问询回复、梳理回复思路或组织回复材料时触发此技能。适用于投行、律所、会所等中介机构的再融资项目场景。
.claude/skills/thomasmoreai-regulatory-inquiry-response-refinancing/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 64% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 58% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 68% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 53% | 0% |
使用此技能当：
首先向用户确认以下信息：
将问询问题按以下类型分类：
为每个问题生成回复框架：
markdown## 问题 [X]：[问题标题] ### 问题原文 [交易所/证监会问询原文] ### 回复要点 1. [核心观点 1] 2. [核心观点 2] 3. [核心观点 3] ### 回复正文 #### 一、[分论点 1] [详细论述 + 数据支撑] #### 二、[分论点 2] [详细论述 + 数据支撑] ### 核查程序 1. [核查程序 1] 2. [核查程序 2] ### 核查结论 [明确结论] ### 附件材料 - [附件 1 名称] - [附件 2 名称]
| 问题类型 | 回复策略 | |----------|----------| | 融资必要性不足 | 详细测算资金缺口、展示项目紧迫性 | | 募投项目前景不明 | 提供市场调研、订单意向、可行性研究 | | 摊薄影响较大 | 说明长期价值、填补回报措施 | | 前次募资使用缓慢 | 解释原因、展示改进措施、调整计划 | | 发行对象关联关系 | 披露关联关系、说明定价公允性 |
markdown## 问题 [X] 核查材料清单 ### 募投项目相关文件 - [ ] 项目可行性研究报告 - [ ] 立项批复文件 - [ ] 环评批复 - [ ] 用地预审意见 ### 财务测算文件 - [ ] 投资估算明细 - [ ] 效益测算过程 - [ ] 敏感性分析 ### 合规性文件 - [ ] 董事会/股东大会决议 - [ ] 前次募资使用情况报告 - [ ] 填补回报措施承诺
references/refinancing-inquiry-types.md - 再融资问询问题类型库references/response-templates-refinancing.md - 再融资回复模板references/raising-rules.md - 再融资规则汇编| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 23,994 | 25,087 | +5% | 1 | 1 | 0% | 3,734 | 5,451 | +46% | 0 | 0 | — |
case-02 | fail→fail | 24,829 | 24,903 | +0% | 1 | 1 | 0% | 3,588 | 4,823 | +34% | 0 | 0 | — |
case-03 | fail→fail | 24,055 | 24,845 | +3% | 1 | 1 | 0% | 3,807 | 5,285 | +39% | 0 | 0 | — |
case-04 | pass→pass | 14,727 | 16,012 | +9% | 1 | 1 | 0% | 2,402 | 3,784 | +58% | 0 | 0 | — |
case-05 | pass→pass | 16,956 | 20,320 | +20% | 1 | 1 | 0% | 2,450 | 4,107 | +68% | 0 | 0 | — |
case-06 | pass→pass | 19,073 | 15,881 | -17% | 1 | 1 | 0% | 2,870 | 4,404 | +53% | 0 | 0 | — |
case-07 | pass→pass | 15,206 | 17,521 | +15% | 1 | 1 | 0% | 2,305 | 3,930 | +70% | 0 | 0 | — |
case-08 | fail→pass | 16,813 | 19,357 | +15% | 1 | 1 | 0% | 2,441 | 3,997 | +64% | 0 | 0 | — |
case-09 | pass→pass | 18,621 | 19,811 | +6% | 1 | 1 | 0% | 2,893 | 4,402 | +52% | 0 | 0 | — |
case-10 | fail→fail | 19,055 | 20,608 | +8% | 1 | 1 | 0% | 3,041 | 4,578 | +51% | 0 | 0 | — |
case-11 | fail→pass | 12,012 | 8,959 | -25% | 1 | 1 | 0% | 1,859 | 2,768 | +49% | 0 | 0 | — |
case-12 | pass→pass | 10,386 | 12,236 | +18% | 1 | 1 | 0% | 1,509 | 3,053 | +102% | 0 | 0 | — |
case-13 | pass→pass | 9,189 | 10,360 | +13% | 1 | 1 | 0% | 1,427 | 2,961 | +107% | 0 | 0 | — |
case-14 | pass→pass | 13,963 | 18,676 | +34% | 1 | 1 | 0% | 2,182 | 3,445 | +58% | 0 | 0 | — |
case-15 | pass→pass | 9,524 | 10,781 | +13% | 1 | 1 | 0% | 1,485 | 3,124 | +110% | 0 | 0 | — |
case-16 | pass→pass | 8,943 | 10,581 | +18% | 1 | 1 | 0% | 1,410 | 2,914 | +107% | 0 | 0 | — |
case-17 | pass→pass | 13,990 | 13,174 | -6% | 1 | 1 | 0% | 1,820 | 3,211 | +76% | 0 | 0 | — |
case-18 | pass→pass | 8,634 | 12,456 | +44% | 1 | 1 | 0% | 1,414 | 3,208 | +127% | 0 | 0 | — |
case-19 | pass→pass | 16,706 | 13,998 | -16% | 1 | 1 | 0% | 2,329 | 3,362 | +44% | 0 | 0 | — |
case-20 | pass→pass | 27,343 | 22,679 | -17% | 1 | 1 | 0% | 4,274 | 5,056 | +18% | 0 | 0 | — |
case-21 | pass→pass | 19,190 | 22,018 | +15% | 1 | 1 | 0% | 3,232 | 4,862 | +50% | 0 | 0 | — |
case-22 | pass→pass | 19,961 | 20,286 | +2% | 1 | 1 | 0% | 3,205 | 4,470 | +39% | 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 +9 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.