Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A 股官方数据，来自同花顺（HiThink）官方 API——涨停股票池、连板天梯、龙虎榜、个股异动原因、热榜、A 股官方口径财报（利润表/资产负债表/现金流量表/财务指标）、行情快照与日 K。
.claude/skills/kansoku-trade-hithink-a-share/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 143% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 28% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 17% | 0% |
> 对话回复语言：中文白话（本仓库规则，见根 CLAUDE.md）。
Longbridge 不覆盖 A 股的短线情绪与官方口径财报数据，这个 skill 补这块盲区。触发词：
600519.SH）不要用这个 skill 做什么：
interval=1d），不支持分钟级。thscode 形如 600519.SH / 000001.SZ / 430047.BJ）。需要 HITHINK_FINANCE_API_KEY（已在仓库根 .env 里配置，脚本通过 _shared/env.py 自动加载，无需手动 source）。缺失时脚本报 exit_code=2，提示去 .env 里加。
| 脚本 | 用途 | 关键参数 | | --------------- | -------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------- | | snapshot.py | 行情快照（多 thscode 批量，或全市场分页） | thscodes...、--limit/--offset（全市场模式） | | kline.py | 日 K 线 | thscode、--days（便捷窗口）或 --start/--end（毫秒）、--adjust | | special.py | 涨停/连板/龙虎榜/异动/热榜，用 --kind 切换 | --kind limit-up-pool\|limit-up-ladder\|dragon-tiger\|skyrocket\|hot\|hot-history\|hot-rank-trend\|anomaly\|anomaly-stock | | financials.py | 官方口径财报 + 财务指标 | thscode、--kind income\|balance\|cashflow\|indicators、--period annual\|quarterly、--limit 或 --start/--end；indicators 用 --report yyyy-N | | calendar.py | A 股近一年交易日序列 | 无参数 | | search.py | 标的检索 / 批量代码表 | query（thscode/代码/中英文名），--list 切到批量代码表 |
所有脚本都支持 --help / --smoke / --verbose / --json（占位，输出恒为 JSON）/ --fresh（绕过缓存）。缓存在 ~/.cache/market-intel/hithink/，节流默认 ≥0.5 秒/请求（_shared/client.py 的节流机制）。
bash# 行情快照：单只/多只 python3 .claude/skills/hithink-a-share/scripts/snapshot.py 600519.SH python3 .claude/skills/hithink-a-share/scripts/snapshot.py 600519.SH 000001.SZ # 日 K：默认近 120 个自然日，前复权 python3 .claude/skills/hithink-a-share/scripts/kline.py 600519.SH --days 90 python3 .claude/skills/hithink-a-share/scripts/kline.py 000001.SZ --adjust backward # 涨停股票池 / 连板天梯 python3 .claude/skills/hithink-a-share/scripts/special.py --kind limit-up-pool --sort-field limit_up_time --sort-dir desc python3 .claude/skills/hithink-a-share/scripts/special.py --kind limit-up-ladder # 龙虎榜（默认最近交易日，全部榜；可选机构/游资榜 + 指定日期） python3 .claude/skills/hithink-a-share/scripts/special.py --kind dragon-tiger python3 .claude/skills/hithink-a-share/scripts/special.py --kind dragon-tiger --board-type hot_money --date 2026-07-14 # 个股异动原因：当日全部，或按标的批量查 python3 .claude/skills/hithink-a-share/scripts/special.py --kind anomaly --tag-codes LIMIT_UP,SHARP_FALL python3 .claude/skills/hithink-a-share/scripts/special.py --kind anomaly-stock --thscodes 600519.SH,000001.SZ # 热榜：飙升榜 / 热股榜 / 历史热股 / 个股排名走势 python3 .claude/skills/hithink-a-share/scripts/special.py --kind skyrocket --period hour python3 .claude/skills/hithink-a-share/scripts/special.py --kind hot python3 .claude/skills/hithink-a-share/scripts/special.py --kind hot-history --date 2026-07-14 python3 .claude/skills/hithink-a-share/scripts/special.py --kind hot-rank-trend --thscode 300034.SZ --start-date 2026-06-21 --end-date 2026-07-14 # 财报：最近 N 期 / 时间区间二选一 python3 .claude/skills/hithink-a-share/scripts/financials.py 600519.SH --kind income --period annual --limit 3 python3 .claude/skills/hithink-a-share/scripts/financials.py 600519.SH --kind cashflow --period annual --start 1577808000000 --end 1735574400000 python3 .claude/skills/hithink-a-share/scripts/financials.py 300033.SZ --kind indicators --report 2025-1 # 交易日历 python3 .claude/skills/hithink-a-share/scripts/calendar.py # 标的检索 / 批量代码表 python3 .claude/skills/hithink-a-share/scripts/search.py 贵州茅台 python3 .claude/skills/hithink-a-share/scripts/search.py 600519.SH python3 .claude/skills/hithink-a-share/scripts/search.py --list --asset-type a-share-index --limit 500 # 绕过缓存 python3 .claude/skills/hithink-a-share/scripts/snapshot.py 600519.SH --fresh
统一走 _shared/client.py 的信封：
json{"ok": true, "data": [...], "meta": {...}}
失败时：
json{ "error": "hithink error 1003: ...", "hint": "request_id=...", "ok": false }
financials.py 的 --limit 与 --start/--end 互斥，同传或半开区间（只传一个）都会报错——脚本已提前拦截，服务端也会返回 code=1004。indicators（财务指标）走完全不同的参数契约：只认 thscode + report（yyyy-N，N=1/2/3/4 对应一季报/中报/三季报/年报），不支持 period/limit。date 必须是真实交易日，传非交易日直接报错，不会自动回退；省略时服务端自动取最近可用交易日。收盘不久后当日龙虎榜可能还没发布，返回空列表是正常现象，不是接口坏了。[start, end] 跨度不超过 10 年；要多只标的必须分开请求。value 是字符串，不是 number——服务端故意保留原始精度，展示时按 index_id 名称判断单位（含"率""比率"的按百分比，"周转率"按次，"倍数"按倍）。anomaly-analysis-list / anomaly-analysis-stock 的解读文案是同花顺二手归纳，不能替代公司公告/交易所披露作为结论依据，只能当情绪信号参考。| exit code | 含义 | 处理 | | --------- | ------------------------------------------------ | -------------------------------------------------------------------------------------------------- | | 0 | 成功 | 解析 data | | 1 | 参数校验失败（本地拦截，未发请求） | 按 hint 补齐参数 | | 2 | 缺 HITHINK_FINANCE_API_KEY | 提示用户去 .env 补 key | | 3 | HiThink 业务错误（code!=0）或 HTTP 4xx/非 JSON | 参考 error 里的 code 含义（1001 缺参/1002 格式错/1003 越界/1004 参数冲突/2003 无权限/4001 超频） | | 4 | 网络错误 | 建议重试 | | 5 | HTTP 401 | 检查 API Key 是否失效 |
longbridge-* —— 美股 / 港股实时行情、分钟级 K 线、账户持仓。stock-deep-dive —— 单只标的多维度首次尽调（美股）。trading-discipline —— 判读纪律共享源头，TD-LANG-03（只做美股，A 股是明确例外场景）、TD-SOURCE-01（一手信源）等规则见此。| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 16,449 | 6,507 | -60% | 1 | 1 | 0% | 2,514 | 3,148 | +25% | 0 | 0 | — |
case-02 | fail→fail | 16,576 | 8,354 | -50% | 1 | 1 | 0% | 2,372 | 3,223 | +36% | 0 | 0 | — |
case-03 | fail→fail | 19,464 | 7,967 | -59% | 1 | 1 | 0% | 3,072 | 3,111 | +1% | 0 | 0 | — |
case-04 | fail→pass | 11,416 | 18,657 | +63% | 1 | 1 | 0% | 2,052 | 4,982 | +143% | 0 | 0 | — |
case-05 | fail→pass | 24,873 | 7,046 | -72% | 1 | 1 | 0% | 3,661 | 4,018 | +10% | 0 | 0 | — |
case-06 | pass→fail | 19,667 | 11,416 | -42% | 1 | 1 | 0% | 2,825 | 3,683 | +30% | 0 | 0 | — |
case-07 | fail→pass | 18,789 | 3,999 | -79% | 1 | 1 | 0% | 2,659 | 3,445 | +30% | 0 | 0 | — |
case-08 | pass→pass | 9,014 | 5,734 | -36% | 1 | 1 | 0% | 1,436 | 3,465 | +141% | 0 | 0 | — |
case-09 | pass→pass | 15,913 | 7,700 | -52% | 1 | 1 | 0% | 1,869 | 3,805 | +104% | 0 | 0 | — |
case-10 | fail→pass | 21,480 | 4,673 | -78% | 1 | 1 | 0% | 2,747 | 3,524 | +28% | 0 | 0 | — |
case-11 | fail→fail | 19,247 | 7,300 | -62% | 1 | 1 | 0% | 3,336 | 3,245 | -3% | 0 | 0 | — |
case-12 | fail→fail | 12,698 | 7,376 | -42% | 1 | 1 | 0% | 1,509 | 3,168 | +110% | 0 | 0 | — |
case-13 | fail→pass | 18,172 | 4,553 | -75% | 1 | 1 | 0% | 2,952 | 3,442 | +17% | 0 | 0 | — |
case-14 | fail→pass | 12,547 | 4,645 | -63% | 1 | 1 | 0% | 2,137 | 3,285 | +54% | 0 | 0 | — |
case-15 | fail→fail | 17,767 | 7,835 | -56% | 1 | 1 | 0% | 3,146 | 3,057 | -3% | 0 | 0 | — |
case-16 | fail→pass | 14,640 | 4,412 | -70% | 1 | 1 | 0% | 2,310 | 3,497 | +51% | 0 | 0 | — |
case-17 | fail→pass | 24,927 | 6,026 | -76% | 1 | 1 | 0% | 3,281 | 3,965 | +21% | 0 | 0 | — |
case-18 | fail→fail | 14,321 | 5,042 | -65% | 1 | 1 | 0% | 2,464 | 2,998 | +22% | 0 | 0 | — |
case-19 | pass→pass | 13,900 | 7,588 | -45% | 1 | 1 | 0% | 2,353 | 3,839 | +63% | 0 | 0 | — |
case-20 | fail→pass | 8,655 | 4,376 | -49% | 1 | 1 | 0% | 1,444 | 3,362 | +133% | 0 | 0 | — |
case-21 | fail→pass | 18,391 | 5,644 | -69% | 1 | 1 | 0% | 3,081 | 3,544 | +15% | 0 | 0 | — |
case-22 | fail→pass | 15,102 | 8,757 | -42% | 1 | 1 | 0% | 2,079 | 3,978 | +91% | 0 | 0 | — |
case-23 | fail→pass | 20,628 | 15,044 | -27% | 1 | 1 | 0% | 3,397 | 4,839 | +42% | 0 | 0 | — |
case-24 | fail→pass | 20,448 | 5,540 | -73% | 1 | 1 | 0% | 3,147 | 3,485 | +11% | 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. 24 cases were attempted, and 16 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 +50 percentage points is the difference between those two pass rates over the 16 comparable cases. 1 case got worse with the skill loaded, and it is 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.