a-shares-master

name	a-shares-master
description	Analyze A-share trades with latest trusted online data.
version	2.1.0
author	tinylion1024
license	MIT
platforms	["linux","macos"]
homepage	https://github.com/tinylion1024/a-share-profit-helper
user-invocable	true
tags	["finance","stocks","a-share","trading"]
category	research
metadata	{"openclaw":{"emoji":"📈","homepage":"https://github.com/tinylion1024/a-share-profit-helper","skillKey":"a-shares-master","os":["linux","macos"],"requires":{"bins":["python3"]},"envVars":[{"name":"A_SHARE_SKILL_OFFLINE_MODE","required":false,"description":"Force fixture mode only for tests or incident fallback."},{"name":"A_SHARE_SKILL_DATA_PATH","required":false,"description":"Path to fixture JSON with market and stock data."},{"name":"A_SHARE_SKILL_LIVE_SOURCE","required":false,"description":"Current trusted upstream, default is tencent."},{"name":"A_SHARE_SKILL_WATCHLIST","required":false,"description":"Comma-separated online watchlist for live stock picking."},{"name":"A_SHARE_SKILL_TIMEOUT_SECONDS","required":false,"description":"HTTP timeout for live upstream requests."},{"name":"IWENCAI_BASE_URL","required":false,"description":"Override iwencai API base URL for semantic search."},{"name":"IWENCAI_API_KEY","required":false,"description":"Enable live iwencai semantic and structured queries."}]},"hermes":{"category":"research","tags":["finance","stocks","a-share","trading"],"related_skills":["browser","spreadsheets"],"config":{"offline_mode":"A_SHARE_SKILL_OFFLINE_MODE","sample_data_path":"A_SHARE_SKILL_DATA_PATH","live_source":"A_SHARE_SKILL_LIVE_SOURCE","watchlist":"A_SHARE_SKILL_WATCHLIST","iwencai_api_key":"IWENCAI_API_KEY"}}}

A-Shares Master Skill

This skill uses the latest trusted online A-share data by default, then converts it into 诊股 / 选股 / 风控 / 盘前盘后复盘 / 市场周期判断 / 方法论执行手册 / 新闻 / 公告 / 资金流 / 行业轮动 / 热点强势股 / 淘股吧热点 / 淘股吧舆情 / 淘股吧个股情绪 / 淘股吧大V观点 / 概念板块 / 研报 / 龙虎榜 / 融资融券 / 大宗交易 / 股东户数 / 分红 / 解禁 / 北向资金 / 个股信息 / 批量实时行情 / K线 / 五档盘口 / 逐笔成交 / 季报快照 / F10 公司资料 / 财报三表 / 一致预期 / 估值 / 批量对比 / 主题研报批量检索 / 快速调研 / iwencai 检索 output. Local fixtures remain available only for tests or controlled fallback.

The six flagship workflows are now market-cycle, leaders, diagnose, playbook, quick-research, and theme-research. They are the preferred external entrypoints. valuation, reports, fund-flow, taoguba-*, and the rest remain available as supporting tools behind those flagship workflows. self-check also reports dependency state, credential state, capability checks, flagship workflows, and health summary.

When To Use

• The user asks whether a stock can be bought, held, reduced, or avoided.
• The user wants a pre-market view, post-market review, or a two-scenario trading plan.
• The agent needs current A-share quotes, turnover, MA20, support, and resistance from live upstreams.
• The user wants stock news, CNInfo announcements, intraday fund flow, sector rotation, THS hot stocks, concept tags, Eastmoney research reports, dragon-tiger seats, margin data, block trades, holder changes, dividend history, northbound flow, stock basics, batch quotes, K-line bars, five-level order book, tick transactions, quarterly snapshots, F10 company text, financial statements, THS consensus EPS, valuation comparison, thematic report research, or iwencai semantic search.
• The user wants Taoguba forum hotspots, market sentiment, retail mood for a stock, or curated VIP viewpoints.
• The agent must use terminal with a single, stable entrypoint that works in both OpenClaw and Hermes-style environments.

Prerequisites

• Run inside {baseDir} or the repository root.
• Python 3.9+ is enough.
• Network access is required for default execution because the skill fetches online quotes and K-lines.
• quarterly-snapshot and f10 use optional mootdx; install it with pip install .[mootdx].
• To force fixture mode for tests or incident fallback, set A_SHARE_SKILL_OFFLINE_MODE=true.

How To Run

Prefer the unified entrypoint:

cd {baseDir}
pip install .
a-shares-skill self-check
a-shares-skill valuation --code 300750

# Repository mode
python3 scripts/run_skill.py self-check
python3 scripts/run_skill.py flagship
python3 scripts/run_skill.py risk --code 300750
python3 scripts/run_skill.py diagnose --code 300750 --date 2026-05-28
python3 scripts/run_skill.py pick --filters basic,tech,catalyst
python3 scripts/run_skill.py market-cycle --date 2026-05-28
python3 scripts/run_skill.py leaders --date 2026-05-28
python3 scripts/run_skill.py pre-market --date 2026-05-28
python3 scripts/run_skill.py post-market --date 2026-05-28
python3 scripts/run_skill.py plan --code 300750 --date 2026-05-28
python3 scripts/run_skill.py playbook --code 300750 --date 2026-05-28
python3 scripts/run_skill.py review-trade --code 300750 --outcome win --return-pct 8.5 --holding-days 3 --theme 储能
python3 scripts/run_skill.py weekly-review --limit 10
python3 scripts/run_skill.py memory-feedback --limit 10
python3 scripts/run_skill.py news --code 300750 --page-size 3
python3 scripts/run_skill.py telegraph --page-size 5
python3 scripts/run_skill.py global-news --page-size 5
python3 scripts/run_skill.py announcements --code 300750 --page-size 3
python3 scripts/run_skill.py fund-flow --code 300750 --period minute
python3 scripts/run_skill.py sectors --top 5
python3 scripts/run_skill.py hot-stocks --page-size 5
python3 scripts/run_skill.py taoguba-hot --page-size 5
python3 scripts/run_skill.py taoguba-sentiment --page-size 10
python3 scripts/run_skill.py taoguba-stock --code 300750 --page-size 10
python3 scripts/run_skill.py taoguba-vip --code 300750 --page-size 5
python3 scripts/run_skill.py concept-blocks --code 300750
python3 scripts/run_skill.py reports --code 300750 --page-size 3
python3 scripts/run_skill.py dragon-tiger --code 002428 --date 2026-05-28
python3 scripts/run_skill.py daily-dragon-tiger --date 2026-05-28 --min-net-buy 5000
python3 scripts/run_skill.py margin --code 300750 --page-size 3
python3 scripts/run_skill.py block-trades --code 300750 --page-size 3
python3 scripts/run_skill.py holders --code 300750 --page-size 3
python3 scripts/run_skill.py dividends --code 300750 --page-size 3
python3 scripts/run_skill.py lockup --code 002428 --date 2026-05-28 --forward-days 90
python3 scripts/run_skill.py northbound --history-days 5
python3 scripts/run_skill.py stock-info --code 300750
python3 scripts/run_skill.py quotes --codes 300750,sh000300,510300 --kind auto
python3 scripts/run_skill.py valuation --code 300750
python3 scripts/run_skill.py compare --codes 300750,002594
python3 scripts/run_skill.py theme-research --queries "人形机器人产业链深度 2026,人形机器人减速器 丝杠" --channel report --size 5 --supplement-per-stock 2
python3 scripts/run_skill.py quick-research --code 300750 --date 2026-05-29
python3 scripts/run_skill.py kline --code 300750 --frequency 4 --limit 5
python3 scripts/run_skill.py order-book --code 300750
python3 scripts/run_skill.py transactions --code 300750 --start 0 --limit 5
python3 scripts/run_skill.py quarterly-snapshot --code 300750
python3 scripts/run_skill.py f10 --code 300750 --category 最新提示
python3 scripts/run_skill.py finance --code 300750 --report-type lrb --page-size 1
python3 scripts/run_skill.py consensus-eps --code 300750
python3 scripts/run_skill.py iwencai-search --query "人形机器人 行星滚柱丝杠 2026" --channel report --size 3

For machine-readable output:

python3 scripts/run_skill.py --format json risk --code 300750

Quick Reference

• terminal Run all commands in this skill through the terminal tool or a shell-capable agent step.
• scripts/run_skill.py Single agent-friendly entrypoint.
• a-shares-skill Installed console script for non-repo usage.
• python3 -m src.main Module entrypoint that mirrors the installed console script.
• src/skill.py Unified facade used by all scripts.
• src/providers/offline.py Fixture provider kept for tests and incident fallback only.
• src/providers/live.py Trusted live provider using Tencent finance plus Eastmoney, THS, mootdx, CNInfo, and CLS endpoints.
• src/core/ Analysis, rating, risk, and workflow logic.
• scripts/run_tests.sh Runs pytest when available, otherwise falls back to unittest.

Natural Language Routing

• In OpenClaw, do not ask the user to type shell commands. The user should speak in natural language and the agent should run the matching command internally.
• Treat Chinese stock short names and six-digit stock codes as equivalent inputs for stock-specific workflows. If the user says 宁德时代能买吗, 比亚迪今天怎么看, or 查一下贵州茅台的 F10, pass the short name directly and let the skill resolve it to a stock code.
• If the user asks a broad question with no command words, route by intent rather than by literal keyword matching.
• If a request mixes market view plus single-stock execution, run the market workflow first, then the stock workflow, and summarize them together.

Examples:

• 今天市场情绪怎么样，适合进攻还是防守？ Run market-cycle.
• 今天主线和龙头是谁？ Run leaders.
• 宁德时代能买吗？给我买卖建议。 Run diagnose, and use playbook if the user also wants entry/add/reduce/exit rules.
• 帮我看一下比亚迪风险大不大。 Run risk.
• 淘股吧今天最热在聊什么？ Run taoguba-sentiment, and use taoguba-hot if the user wants the underlying posts.
• 宁德时代在淘股吧的情绪怎么样，有没有大V观点？ Run taoguba-stock, and use taoguba-vip if the user wants more VIP posts.
• 今天有哪些值得看的票？ Run pick, optionally followed by hot-stocks or sectors when the user asks for why they are strong.
• 查一下宁德时代最近公告、资金流和研报。 Run announcements, fund-flow, and reports.
• 机器人主题最近有什么研报和标的？ Run theme-research.
• 帮我快速研究一下寒武纪。 Run quick-research.
• 帮我复盘最近交易，看看我适合什么 setup。 Run weekly-review, then memory-feedback.

Procedure

Short/Mid-Term Selection Heuristic

When the user asks 谁更有机会 / 帮我选股 / 高盈亏比 / 舒服买点 for a holding period around 2 to 10 trading days, rank candidates with this priority:

1. Theme strength and catalyst clarity first Prefer names that belong to the current main line, or directly benefit from an active expansion theme.
2. Trend continuity and overhead supply second Prefer stocks already walking in a clean uptrend, or trend mid-caps that just completed a healthy pullback and regained support.
3. Volume-price structure and comfort third Prefer pullback support, breakout retest, or 缩量整理后的放量再启动. Avoid consensus 高潮日、连续大涨后的追高 and single-day weak-to-strong names with heavy overhead supply.
4. Static valuation only after the first three pass Lower PE/PEG alone must not outrank a stronger trend leader. Cheap laggards stay on the watchlist unless they also get a fresh catalyst and a visible trend repair.

Use these labels explicitly when helpful:

• 趋势票 Main-line trend, suitable for pullback or confirmation entries.
• 性价比票 Valuation is better, but timing depends on catalyst rotation or trend repair.
• 修复观察票 Logic exists, but structure is still weak; do not rank it ahead of cleaner leaders.

If the user asks for a comfortable buy point, default to:

• support retest near key moving averages or prior breakout zones
• confirmation after reclaiming an important intraday or daily level
• no recommendation to chase a consensus acceleration unless the user explicitly wants breakout momentum trades

1. Run python3 scripts/run_skill.py self-check before using live scenarios.
2. Prefer the flagship workflows first: market-cycle, leaders, diagnose, playbook, quick-research, theme-research.
3. If the user asks for 风控 only, call risk.
4. If the user asks for 买/卖/持建议, call diagnose.
5. If the user asks for what to buy, call pick.
6. If the user asks for市场情绪阶段、仓位节奏 or 主攻/防守判断, call market-cycle.
7. If the user asks for主线、龙头 or 题材梯队, call leaders.
8. If the user needs single-stock entry/add/reduce/exit rules, call playbook.
9. If the user needs actionable prices and position sizing, call plan.
10. If the user asks for event flow, call news, telegraph, global-news, or announcements.
11. If the user asks for资金面 or 行业轮动, call fund-flow or sectors.
12. If the user asks for强势股归因 or 题材热点, call hot-stocks.
13. If the user asks for淘股吧热点、大V观点、市场舆情 or 个股股民情绪, call taoguba-hot, taoguba-sentiment, taoguba-stock, or taoguba-vip.
14. If the user asks for个股所属概念、概念标签 or 题材归属, call concept-blocks.
15. If the user asks for卖方观点 or 机构覆盖, call reports.
16. If the user asks for席位资金、连板观察 or 全市场龙虎榜, call dragon-tiger or daily-dragon-tiger.
17. If the user asks for两融、筹码集中、大宗成交、分红回报 or 解禁预警, call margin, holders, block-trades, dividends, or lockup.
18. If the user asks for指数、ETF or 批量实时行情, call quotes.
19. If the user asks for个股估值、PEG、PE消化 or 批量估值对比, call valuation or compare.
20. If the user asks for主题研报聚合、跨主题研报批量检索 or 主题补充研报, call theme-research.
21. If the user asks for新标的快速调研, call quick-research.
22. If the user asks for交易后复盘, call review-trade.
23. If the user asks for周期复盘总结, call weekly-review.
24. If the user asks for把复盘结果转成下一阶段建议, call memory-feedback.
25. If the user asks for个股K线、五档盘口 or 逐笔成交, call kline, order-book, or transactions.
26. If the user asks for北向资金、个股基本面、季报快照、F10资料、财务报表 or 一致预期, call northbound, stock-info, quarterly-snapshot, f10, finance, or consensus-eps.
27. If the user asks for跨主题语义检索 or 结构化问财查询, call iwencai-search or iwencai-query.
28. Use --format json when another agent or tool will parse the output.
29. If live data is unavailable, fail clearly unless the caller explicitly enabled offline fallback.

Pitfalls

• Do not assume fixture output is acceptable in production; latest online data is the default contract.
• Do not call the legacy scripts first; they are thin wrappers around run_skill.py.
• Do not hide upstream failures by silently switching to fixture data.
• Do not edit references/ to change behavior. The executable behavior lives in src/ and scripts/.

Verification

• Run python3 scripts/run_skill.py self-check.
• Run python3 -m src.main --format json self-check.
• Run sh scripts/run_tests.sh.
• If pytest is installed, the same tests should also pass via python3 -m pytest -q tests.