Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股困境反转/业绩拐点分析。当用户说"困境反转"、"拐点"、"turnaround"、"触底反弹"、"业绩拐点"、"什么时候好转"、"XX跌了很多还能买吗"、"底部反转"、"周期底部"、"扭亏"时触发。评估处于低谷的企业是否具备反转条件，分析反转驱动因素（行业周期/管理层变革/产品创新/政策利好），寻找业绩拐点信号。支持研报风格（formal）和快速判断风格（brief）。
.claude/skills/aifinlab-a-share-turnaround/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | 36% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 63% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 33% | 0% |
bashSCRIPTS="$SKILLS_ROOT/cn-stock-data/scripts" # 多期财务数据（至少8季度，对比业绩下滑→企稳→反转趋势） python "$SCRIPTS/cn_stock_data.py" finance --code [CODE] # 最新行情（当前估值水平、距高点回撤幅度） python "$SCRIPTS/cn_stock_data.py" quote --code [CODE] # 长周期 K 线（2-3 年，定位当前在周期中的位置） python "$SCRIPTS/cn_stock_data.py" kline --code [CODE] --freq daily --start [2年前日期] # 资金流向（是否有聪明资金提前进场） python "$SCRIPTS/cn_stock_data.py" fund_flow --code [CODE] --days 60
补充：通过 web 搜索获取：
确认"困境"的性质和原因，这是判断能否反转的基础：
困境类型分类： | 类型 | 特征 | 反转可能性 | 典型案例 | |------|------|-----------|---------| | 行业周期性低谷 | 全行业低迷，非公司个体问题 | 高（周期必然回归） | 猪周期/芯片周期/航运周期 | | 公司经营失误 | 战略失误/扩张过快/管理混乱 | 中（需要换帅或重组） | 盲目多元化/产品失败 | | 一次性冲击 | 疫情/自然灾害/黑天鹅 | 高（冲击消退后自然恢复） | 旅游/餐饮受疫情冲击 | | 结构性衰退 | 技术替代/需求消亡 | 低（除非成功转型） | 功能机→智能机/纸媒 | | 财务造假/暴雷 | 历史财务不可信 | 极低（信用已破产） | ST 公司 |
需收集的数据：
逐项检查是否出现反转信号（参考 references/turnaround-guide.md）：
财务信号（从财报数据中提取）：
运营信号（通过 web 搜索获取）：
管理信号：
外部信号：
同行业公司反转进度对比： | 公司 | 困境程度 | 反转信号数量 | 最新业绩趋势 | 股价表现 | |------|---------|------------|------------|---------| | 目标公司 | | | | | | 同行A | | | | | | 同行B | | | | |
关键判断：
当前估值 vs 历史分位： | 指标 | 当前值 | 历史最低 | 历史中位 | 反转后合理值 | |------|-------|---------|---------|------------| | PE(TTM) | | | | | | PB | | | | | | PS | | | | | | EV/EBITDA | | | | |
注意事项：
反转概率评估（综合打分）：
| 维度 | 权重 | 评分(1-5) | 说明 | |------|------|----------|------| | 困境可逆性 | 25% | | 结构性vs周期性 | | 反转信号强度 | 30% | | 已出现信号数量和质量 | | 管理层能力 | 15% | | 是否有改善意愿和执行力 | | 估值安全边际 | 20% | | 下行空间有限 | | 催化剂明确度 | 10% | | 是否有近期可见催化剂 | | 综合得分 | 100% | | 加权总分 |
评分标准：
根据用户需求选择 formal 或 brief 风格，参考 references/turnaround-guide.md 中的框架。
| 维度 | formal（研报风格） | brief（快速判断） | |------|-------------------|-----------------| | 篇幅 | 2-4 页 | 半页 | | 困境诊断 | 完整原因分析+历史复盘 | 一句话概括困境类型 | | 反转信号 | 逐项检查清单+数据支撑 | 只列最关键的 3-5 个信号 | | 同行对比 | 完整表格 | 简述领先/滞后情况 | | 估值分析 | 多指标+情景测算 | PB 分位+一句话结论 | | 概率评估 | 完整打分表 | 直接给结论和概率 | | 操作建议 | 不给具体建议 | 可加个人操作思路 | | 免责声明 | 需要 | 不需要 |
默认风格：brief
python# 调用 skill result = run_skill({ "param1": "value1", "param2": "value2" })
bashpython scripts/run_skill.py --input data.json
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 19,372 | 30,331 | +57% | 1 | 1 | 0% | 3,101 | 2,836 | -9% | 0 | 0 | — |
case-02 | fail→fail | 39,446 | 12,134 | -69% | 1 | 1 | 0% | 5,959 | 2,992 | -50% | 0 | 0 | — |
case-03 | fail→fail | 27,588 | 9,029 | -67% | 1 | 1 | 0% | 3,941 | 3,651 | -7% | 0 | 0 | — |
case-04 | pass→pass | 16,943 | 16,529 | -2% | 1 | 1 | 0% | 2,471 | 4,630 | +87% | 0 | 0 | — |
case-13 | fail→pass | 12,379 | 2,517 | -80% | 1 | 1 | 0% | 1,835 | 2,493 | +36% | 0 | 0 | — |
case-14 | pass→pass | 17,211 | 3,922 | -77% | 1 | 1 | 0% | 2,776 | 2,751 | -1% | 0 | 0 | — |
case-20 | fail→pass | 15,486 | 3,255 | -79% | 1 | 1 | 0% | 2,375 | 2,594 | +9% | 0 | 0 | — |
case-21 | fail→pass | 32,250 | 2,589 | -92% | 1 | 1 | 0% | 1,960 | 2,498 | +27% | 0 | 0 | — |
case-05 | pass→pass | 17,973 | 12,373 | -31% | 1 | 1 | 0% | 2,610 | 3,714 | +42% | 0 | 0 | — |
case-06 | pass→pass | 18,416 | 14,088 | -24% | 1 | 1 | 0% | 2,448 | 3,992 | +63% | 0 | 0 | — |
case-07 | pass→pass | 18,532 | 18,589 | +0% | 1 | 1 | 0% | 2,754 | 5,021 | +82% | 0 | 0 | — |
case-08 | pass→pass | 20,243 | 18,924 | -7% | 1 | 1 | 0% | 3,049 | 4,949 | +62% | 0 | 0 | — |
case-09 | pass→pass | 19,940 | 20,253 | +2% | 1 | 1 | 0% | 2,737 | 4,925 | +80% | 0 | 0 | — |
case-10 | pass→pass | 18,171 | 14,812 | -18% | 1 | 1 | 0% | 2,572 | 4,310 | +68% | 0 | 0 | — |
case-11 | pass→pass | 19,359 | 19,769 | +2% | 1 | 1 | 0% | 2,759 | 4,940 | +79% | 0 | 0 | — |
case-12 | pass→pass | 6,702 | 4,764 | -29% | 1 | 1 | 0% | 1,324 | 2,814 | +113% | 0 | 0 | — |
case-15 | fail→pass | 19,377 | 11,043 | -43% | 1 | 1 | 0% | 2,240 | 3,642 | +63% | 0 | 0 | — |
case-16 | fail→pass | 11,912 | 2,890 | -76% | 1 | 1 | 0% | 2,010 | 2,682 | +33% | 0 | 0 | — |
case-17 | pass→pass | 13,947 | 5,610 | -60% | 1 | 1 | 0% | 1,920 | 2,829 | +47% | 0 | 0 | — |
case-18 | pass→pass | 19,095 | 16,704 | -13% | 1 | 1 | 0% | 2,759 | 4,694 | +70% | 0 | 0 | — |
case-19 | fail→pass | 16,021 | 4,171 | -74% | 1 | 1 | 0% | 2,382 | 2,730 | +15% | 0 | 0 | — |
case-22 | pass→pass | 22,119 | 19,333 | -13% | 1 | 1 | 0% | 2,865 | 4,917 | +72% | 0 | 0 | — |
case-23 | pass→fail | 18,805 | 10,329 | -45% | 1 | 1 | 0% | 3,017 | 2,618 | -13% | 0 | 0 | — |
case-24 | pass→fail | 21,378 | 53,245 | +149% | 1 | 1 | 0% | 2,853 | 2,494 | -13% | 0 | 0 | — |
case-25 | pass→fail | 29,549 | 92,959 | +215% | 1 | 1 | 0% | 3,577 | 2,795 | -22% | 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. 25 cases were attempted, and 20 counted toward the lift figure. The other 5 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 +12 percentage points is the difference between those two pass rates over the 20 comparable cases. 3 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.