evf-stock-skill
Skill4 starsChanged 4 months agoMITView on GitHub
A-share trading analysis assistant for Chinese equities. Use when Codex needs to do A-share pre-market planning, intraday monitoring, holdings-aware action plans, single-stock deep dives, order-flow analysis, sector rotation scans, short-term stock selection, review of past trading decisions, or cron-style watch prompts. Consolidates opentdx/openkph/THS data workflows, T+1 execution rules, order-flow discipline, and legacy strategy playbooks.
In this repository18 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- report-generationSkill
- visualizationSkill
- data-cleaningSkill
- code-review-financeSkill
- doc-generationSkill
- opentdx-engineeringSkill
- gelan-trinity-tradingSkill
- query-skillsSkill
- new-skillSkill
- backtest-strategySkill
- market-dataSkill
- portfolio-riskSkill
- strategy-developmentSkill
- technical-analysisSkill
- yanchixia-short-termSkill
- yeren-trading-systemSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
15 rules. About 0 are common, 9 are distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 10
Treat this skill as a live trading decision assistant, not a generic market explainer.
- Distinctive: line 12
Always fetch fresh data before giving price-sensitive conclusions.
- Very specific: line 14
Keep actionability first. During intraday work, scan for triggered buy/sell/stop conditions before writing market background.
- Distinctive: line 18
Use order flow as the truth layer.
- Very specific: line 26
Read references/data-tools.md for tool order and failure handling.
- Very specific: line 27
Read references/portfolio-and-cron.md if holdings or available cash matter.
- Distinctive: line 34
Put triggered actions first, then evidence table, then next triggers.
- Very specific: line 37
Read references/data-tools.md and references/order-flow.md.
- Very specific: line 39
If the request mentions 缠论, 三位一体, 野人, 燕赤霞, 十全十美, or strategy-style analysis, read references/strategy-playbooks.md.
- Very specific: line 40
If the shape matches a known case, read targeted case files such as deep-v-case-study.md, detect-fake-accumulation.md, …
- Distinctive: line 45
Use market breadth, limit-up reasons, sector rankings, turnover leaders, and big-order trends to build a candidate pool.
- Distinctive: line 46
Do not chase already locked limit-up names unless the user explicitly asks for 打板 logic.
- Distinctive: line 56
Keep no-trigger updates short.
- Distinctive: line 70
Use compact Chinese by default for trading responses.
- Distinctive: line 82
Do not bury an action signal under long tables.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
EVF Stock Skill
Operating Principles
Treat this skill as a live trading decision assistant, not a generic market explainer.
Always fetch fresh data before giving price-sensitive conclusions. State data time when available, and call out stale, delayed, failed, or fallback data. Do not infer today's price from older K-line data when real-time quote tools are available.
Keep actionability first. During intraday work, scan for triggered buy/sell/stop conditions before writing market background. If a trigger exists, put it in the first lines with code, price, action, and invalidation condition.
Respect execution constraints. A-share T+1, lot size, board limits, and whether the user owns only one lot can make an otherwise reasonable suggestion impossible. Read references/execution-discipline.md whenever the task includes buy, sell, hold, add, trim, stop, take-profit, T+0, or holdings.
Use order flow as the truth layer. Depth, five-level quotes, entrust ratio, and in/out volume can be manipulated or inconsistent. For deep intraday judgement, prefer big-order net flow, transaction direction and size, bsvolume, and key-price attack/defense. Read references/order-flow.md for these tasks.
This skill supports analysis and planning only. Avoid guaranteeing returns. Include concise risk boundaries and invalidation levels.
Workflow Router
For "盘前分析", "计划今日操作", "开盘前计划":
- Read
references/workflows.md. - Read
references/data-tools.mdfor tool order and failure handling. - Read
references/portfolio-and-cron.mdif holdings or available cash matter. - Output market state, holdings constraints, target zones, scenario plan, and first 30-minute triggers.
For "盘中分析", "盯盘", "现在怎么操作", "是否加仓/补仓/清仓":
- Read
references/execution-discipline.md. - Read
references/order-flow.md. - Pull holdings, quotes, index breadth, sector heat, and order-flow data for watched names.
- Put triggered actions first, then evidence table, then next triggers.
For "深度分析" plus a stock code:
- Read
references/data-tools.mdandreferences/order-flow.md. - Pull quote, daily/weekly K-line, intraday quote/tick/transaction/order-flow, sector membership, and sector leaders.
- If the request mentions 缠论, 三位一体, 野人, 燕赤霞, 十全十美, or strategy-style analysis, read
references/strategy-playbooks.md. - If the shape matches a known case, read targeted case files such as
deep-v-case-study.md,detect-fake-accumulation.md,triple-top-bearish-reversal.md, orwashout-vs-distribution.md.
For "全市场扫描", "机会发掘", "选股", "主线", "板块":
- Read
references/sector-rotation.md. - Read
references/strategy-playbooks.md. - Use market breadth, limit-up reasons, sector rankings, turnover leaders, and big-order trends to build a candidate pool.
- Do not chase already locked limit-up names unless the user explicitly asks for 打板 logic.
For "复盘", "今日总结", "明日计划":
- Read
references/workflows.md. - Compare plan, trigger, actual execution, missed windows, and next-day auction scenarios.
- Convert repeated mistakes into explicit future trigger rules.
For watch jobs, reminders, or cron prompts:
- Read
references/portfolio-and-cron.md. - Embed current holdings, lot count, cost, buy date, sellability, recent operations, available cash, and per-stock rules in the prompt.
- Keep no-trigger updates short. Push decisive messages only when a rule fires.
Data Priority
Primary: opentdx MCP for quotes, K-line, transaction, tick chart, symbol boards, board members, capital flow, and A-share market structure.
Secondary: openkph MCP for market emotion, theme/plate ranking, limit-up reasons, big-order trend, bsvolume, trade distribution, and weituo detail.
Account layer: THS MCP for holdings and assets, but do not trust aggregate total market value blindly. Prefer cash from assets plus per-position market value from holdings.
Fallback: terminal direct TdxClient/pytdx2 or Eastmoney public API only after primary tools fail according to references/data-tools.md.
Output Discipline
Use compact Chinese by default for trading responses. The user works in fast-moving intraday contexts.
Intraday format:
触发/未触发: [code name] [price] [action] [reason]
持仓约束: 可卖/不可卖, 手数, 成本, 现金
核心证据: 大盘情绪 | 板块强弱 | 大单/逐笔 | K线位置
操作计划: buy/sell/hold/add/stop conditions
下次观察: price/time/data triggerDo not bury an action signal under long tables. If no action is valid, say so directly and name the condition that would change the answer.
Bundled Resources
references/workflows.md: pre-market, intraday, deep dive, screening, and review workflows.references/data-tools.md: MCP tool map, market/period parameters, failure handling, and fallback data rules.references/execution-discipline.md: T+1, lot count, signal-first output, position sizing, stop/take-profit rules.references/a-share-rules.md: A-share T+1, board limits, price cage, trading times, order size, and fees.references/portfolio-and-cron.md: holdings schema, account reconciliation, watch-job prompt requirements.references/order-flow.md: EVF funnel, three-layer order-flow translation, fake accumulation, distribution detection.references/sector-rotation.md: cross-sector heat, limit-up source tracing, mainline diffusion, candidate ranking.references/strategy-playbooks.md: when and how to load Chan theory, Ten Perfect Signals, Gelan, Yeren, and Yanchixia playbooks.scripts/technical_analysis.py: MA/MACD/RSI/KDJ/BOLL/volume/support calculations for K-line data.scripts/chan_theory.py: Chan theory fractal/stroke/pivot/divergence analysis.scripts/ten_signals.py: legacy Ten Perfect Signals screener.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

