plan-review

name	plan-review
description	Standardized pre-market preparation, noon review, and evening review workflows plus trading journal management. 三段式交易操作系统：盘前准备→午间复盘→晚间复盘，配合交易日志工具记录每一天。 触发词：盘前准备、午间复盘、晚间复盘、交易计划、复盘、写日志、交易日志、今日计划、纪律打分。 ——核心原则：脚本取数据，AI 做分析，交易者负责最终决策。

Plan & Review Skill

交易操作系统：没有计划不交易，没有复盘不成长。

This skill provides the process backbone for the mmtickerlab project. It connects the market data skill and sim-trade execution skill into a disciplined daily workflow.

Architecture principle: Scripts are data faucets — they fetch, aggregate, and structure raw market data. The AI is the analyst — it reads the data, applies reasoning, identifies patterns, and generates the final trading plan. This separation ensures analysis is contextual and adaptive, not hard-coded.

There are three daily checkpoints:

Checkpoint	Who fetches data	Who analyzes
Pre-Market (盘前准备)	pre_market.py	AI synthesizes plan
Noon Review (午间复盘)	noon_review.py	AI checks alignment
Evening Review (晚间复盘)	evening_review.py	AI performs deep复盘

The central artifact is the daily trading journal, stored at data/journal/YYYYMMDD.md.

---

Environment Setup

uv venv skills/plan-review/.venv
uv pip install --python skills/plan-review/.venv -r skills/plan-review/requirements.txt

---

Data Faucet Scripts (all in skills/market/scripts/)

These scripts fetch and structure raw data. They do NOT make final trading decisions — that's the AI's job.

pre_market.py — 盘前数据聚合器

Fetches all pre-market data into structured output. Run first, then AI analyzes.

skills/market/.venv/bin/python skills/market/scripts/pre_market.py [--date YYYYMMDD]

Data provided:

Section	Data
A50期货	最新价 + 涨跌幅
离岸人民币	USD/CNY 买卖报价
美股三大指数	DJIA/SPX/NDX 收盘价 + 涨跌幅
恒生指数	收盘价 + 涨跌幅
宏观新闻	政策/经济相关新闻标题 + 摘要
7大A股指数	收盘价 + 涨跌幅
市场宽度	涨跌家数、涨停/跌停数、成交额
涨停池 Top 15	代码/名称/连板/行业/换手率
概念资金流向 Top 5	净流入/涨幅/领涨股
行业资金流向 Top 5	净流入/涨幅/领涨股
自选股技术指标	K线 + MA/MACD/RSI/BOLL/KDJ（每只）

noon_review.py — 午间数据聚合器

skills/market/.venv/bin/python skills/market/scripts/noon_review.py

Data provided:

• 早盘实时三大指数
• 涨跌分布 + 涨停/跌停/炸板率
• 领涨题材 + 大单成交
• 盘前计划标的实时报价（从 journal 解析）

evening_review.py — 晚间数据聚合器

skills/market/.venv/bin/python skills/market/scripts/evening_review.py [--date YYYYMMDD]

Data provided:

• 全天市场总览（同 pre_market 模块二）
• 主线题材资金流向
• 龙虎榜：机构买卖/游资动向/负反馈标的（stock_lhb_detail_em + stock_lhb_ggtj_sina）
• 连板高度 + 阶梯完整性

---

Journal Script

journal.py (skills/plan-review/scripts/journal.py)

Command	Description
create	Create today's journal (reads plan from --plan or stdin)
append --section {noon,evening}	Append a review section
view [--date YYYYMMDD]	View a journal
list [--n N]	List recent entries

---

Daily Workflows

Phase 1: Pre-Market Preparation (盘前准备)

Goal: Generate a complete, AI-analyzed trading plan before market opens.

Step 1: Fetch Raw Data

skills/market/.venv/bin/python skills/market/scripts/pre_market.py > /tmp/pre_market_data.md

This produces a structured data dump. The AI reads it.

Step 2: AI Analysis & Plan Synthesis

The AI reads the data and performs the following analysis:

模块一：宏观定调

• Read A50 + US indices + RMB + news
• Judge: 今日高开/平开/低开概率？外资情绪偏多/偏空？有无重大政策催化剂？
• Output: 1-2 句宏观定调 + 对仓位的影响

模块二：情绪判断

• Read breadth data (涨跌比、涨停/跌停数、成交额)
• Read 涨停池（连板分布、行业集中度）
• Read 资金流向（概念/行业 Top 5）
• Judge: 赚钱效应在哪里？亏钱效应在哪里？连板梯队是否健康？
• Output: 情绪定性 + 最强 1-2 条主线

模块三：题材筛选

• Cross-reference 资金流向 + 涨停集中行业 + 宏观新闻
• Judge: 哪些题材有持续性？哪些是一日游？
• Output: ≤3 个焦点题材，每个含：逻辑 + 龙头

模块四：交易预案（核心）

• Read 自选股技术指标数据
• For each stock, the AI judges:
  • 当前处于什么技术位置？（趋势/震荡/超跌）
  • 有哪些可识别的入场信号？信号的可靠性如何？
  • 合理的入场价位、止损价位、仓位比例？
• If no clear setups exist, state "今日无符合条件的交易机会" — this is a valid output.
• Output: ≤3 个 If-Then 预案

模块五：仓位纪律

• Based on 模块一+二, AI judges market temperature
• Determines position limit
• Writes today's prohibited behaviors

Step 3: Save to Journal

# AI writes the final plan to a temp file, then:
cat /tmp/final_plan.md | skills/plan-review/.venv/bin/python skills/plan-review/scripts/journal.py create

✅ Pre-Market Completion Criterion

Journal contains AI-generated plan with all 5 modules. No plan = no trading.

---

Phase 2: Noon Review (午间复盘)

Goal: Check if the morning is going according to plan.

Step 1: Fetch Noon Data

skills/market/.venv/bin/python skills/market/scripts/noon_review.py > /tmp/noon_data.md

Step 2: AI Analysis

1. Read the market: 上午指数表现？成交量 vs 昨日同期？领涨/领跌板块？
2. Check against plan: Read today's journal. For each plan stock, check:
   • Did trigger conditions fire?
   • If not, are conditions still valid?
   • Any stock showing unexpected weakness?
3. Self-examination check: Any plan-external impulses?
4. Afternoon strategy: Based on morning patterns, what's the afternoon play?

Step 3: Save

cat /tmp/noon_final.md | skills/plan-review/.venv/bin/python skills/plan-review/scripts/journal.py append --section noon

---

Phase 3: Evening Review (晚间复盘)

Goal: Deep复盘 — market playback, self-scoring, and next-day preparation.

Step 1: Fetch Evening Data

skills/market/.venv/bin/python skills/market/scripts/evening_review.py > /tmp/evening_data.md

This provides: full-day breadth, theme flows, 龙虎榜 data, 连板高度.

Step 2: AI Deep Analysis

① Market Narrative: What was today's dominant story? One sentence.

② Theme Lifecycle: The AI judges each leading theme's stage:

• 启动期 / 发酵期 / 高潮 / 分歧 / 退潮
• This is NOT a formula — the AI considers volume, breadth, news context, sector rotation patterns

③ LHB Analysis: Read 龙虎榜 data and identify:

• Which institutions are buying? What's the pattern?
• Any stocks with asymmetric buy/sell ratios?
• Any warning signals (机构集体出逃)?

④ Candidate Pool: The AI generates 3-5 next-day candidates. For each:

• Why: Concrete reason grounded in today's data (LHB buying + technical position + theme alignment)
• Risk: What could go wrong?
• Watch level: 重点关注 / 一般关注
• This is raw material for tomorrow's pre_market.py, not trading instructions

⑤ Self-Scoring:

• P&L attribution (planned vs unplanned)
• Discipline score (0-100)
• One error logged

Step 3: Save

cat /tmp/evening_final.md | skills/plan-review/.venv/bin/python skills/plan-review/scripts/journal.py append --section evening

✅ Evening Review Completion Criterion

Journal contains: 1) market narrative, 2) theme lifecycle judgment, 3) LHB analysis, 4) candidate pool with reasoning, 5) discipline score + error log.

---

AI Analysis Guidelines

When performing analysis, the AI should:

1. Ground every judgment in data: "Because 涨停 131 家 and 半导体净流入 119 亿..." not "半导体 looks strong"
2. State uncertainty explicitly: "炸板率 59% suggests high分歧 but 131 涨停 still indicates strong underlying demand — the signal is mixed"
3. Quality over quantity: 3 well-analyzed candidates > 10 superficial mentions
4. "No trade" is a valid output: If conditions aren't right, say so clearly
5. Override script suggestions when needed: The scripts provide baseline analysis for efficiency. If the AI disagrees, explain why and provide its own judgment

---

Integration with Existing Skills

Skill	Role
market	Data faucet: pre_market.py, noon_review.py, evening_review.py, plus individual scripts (overview.py, limit_up.py, fund_flow.py, stock_profile.py, news.py)
sim-trade	Execution: portfolio check (pre-market), trade verification (evening)

---

Data Locations

Data	Path
Daily journals	data/journal/YYYYMMDD.md
Output templates	skills/plan-review/templates/*.md
Data scripts	skills/market/scripts/pre_market.py, noon_review.py, evening_review.py
Skill scripts	skills/plan-review/scripts/journal.py
Watchlist	data/watchlist.json