财经周刊展望

写作 社区

技能解读

当前原文的自动摘译,尚未人工审校。

概述

生成 Markdown 每周展望,识别下一个交易周的高概率看涨和看跌行业和股票,涵盖美国股票和中国 A 股。该报告必须结合上游金融文章、新鲜的权威数据、市场行为验证、预期差距推理、情景伪造和明确的交易行动计划。该技能是以下技能的下游:

所需输入

优先选择当前项目的markdown/目录中的最新可用文件: markdown/daily-finance-YYYY-MM-DD.md markdown/finance-core-analysis-YYYY-MM-DD.md markdown/finance-explosive-article-YYYY-MM-DD.md 如果存在多个日期,则使用最近的交易相关日期,除非用户指定日期。保留上游事实和来源,但重新检查最终论文中使用的任何数字。

强制外部数据步骤

在撰写当前的每周展望之前,请务必访问当前的外部数据。尽可能使用主要/权威来源: 官方:中央银行、统计机构、交易所、监管机构、上市公司备案、投资者关系、收益发布 市场数据:交易所页面、ETF/基金发行人数据、CBOE/FRED/Treasury/Nasdaq/NYSE、HKEX、SSE/SZSE、CSRC/PBOC、Wind/Eastmoney/同花顺(仅当官方数据无法获取时)

1. 建立事实基础

从每日/核心/爆炸性文章中提取上游事实,然后添加新数据: 宏观流动性:美联储/央行信号、美国 10 年期、中国 10 年期、DXY、SHIBOR/DR007、财政部/TGA 或美联储资产负债表背景(如果相关) 风险偏好:VIX、看跌/看涨、市场广度、新高/新低、主要指数趋势 资金流向:美国板块 ETF 流向(如果有)、A 股北向/南向(如果有)、保证金融资、主板/科创板/创业板成交额结构

2、确定宏观定位

说明下周是否应该是偏进攻、中性偏进攻、中性偏防守 或偏防守。使用这个链:宏观约束 -> 资金行为 -> 风格偏好 -> 仓位上限 -> 证伪信号 不要说“风险偏好改善”而不命名改善的可观察变量。定位必须包括:

流狐档案 运行与权限由流狐静态检查推断,不代表作者声明或实际测试结果。
领域
写作
作者声明的智能体
未找到明确声明;不据此推断已兼容或已测试
静态检查
88 / 100 · 启发式扫描,不代表运行安全
运行环境证据
未执行;当前没有隔离运行测试证据
作者 / 版本 / 许可证
@digoal · 未声明许可证
上下文用量评级
低消耗
接入复杂程度
需简单配置
是否需要外部接口密钥
不需要
兼容的系统
Windows
运行环境要求
未检出明确要求
检测到的文件与系统行为
  • 只读
  • 允许写入 / 修改
网络行为
仅限本地
安装命令数
无(仅作为资料)

需要注意: 未声明工具权限范围;实际权限由运行环境决定。

作者说明文件(原文) SKILL.md

# Finance Weekly Outlook

## Overview

Generate a Markdown weekly outlook that identifies high-probability bullish and bearish industries and stocks for the next trading week, covering both US equities and China A-shares. The report must combine upstream finance articles, fresh authoritative data, market-behavior validation, expectation-gap reasoning, scenario falsification, and explicit trading action plans.

This skill is downstream of:

`daily-finance -> finance-core-analysis -> finance-explosive-article -> finance-weekly-outlook`

It can also run independently when upstream files are absent, but never invent missing data.

## Required Inputs

Prefer the newest available files in the current project's `markdown/` directory:

1. `markdown/daily-finance-YYYY-MM-DD.md`
2. `markdown/finance-core-analysis-YYYY-MM-DD.md`
3. `markdown/finance-explosive-article-YYYY-MM-DD.md`

If multiple dates exist, use the most recent trading-relevant date unless the user specifies a date. Preserve upstream facts and sources, but re-check any number used in the final thesis.

Resolve the forecast window explicitly:

- For US markets, use the next US trading week.
- For A-shares, use the next China trading week.
- If holidays make the US and China windows different, show both date ranges in the header.
- If the user asks for "next week" during an active session, use the next full trading week unless they ask for the remaining days of the current week.

If upstream files are absent, run independently with fresh data and state `上游文件:未使用` in the report metadata.

## Mandatory External Data Step

Always access current external data before writing a current weekly outlook. Use primary/authoritative sources where possible:

- Official: central banks, statistics agencies, exchanges, regulators, listed-company filings, investor relations, earnings releases
- Market data: exchange pages, ETF/fund issuer data, CBOE/FRED/Treasury/Nasdaq/NYSE, HKEX, SSE/SZSE, CSRC/PBOC, Wind/Eastmoney/同花顺 only when official data is unavailable
- News: Reuters, Bloomberg, FT, WSJ, 财新, 第一财经, 财经, 21世纪经济报道, 经济观察报

Read `references/data-checklist.md` when planning the data pull. Read `references/tradingagents-method.md` when structuring the reasoning and risk review.

Do not rely on self-media, unverified social posts, headlines without source traceability, or a single isolated stock example to support an industry conclusion.

Current data must be no older than the latest completed trading session unless a source publishes less frequently. Label stale indicators and avoid using them as decisive evidence.

## Workflow

### 1. Build the Fact Base

Extract upstream facts from daily/core/explosive articles, then add fresh data:

- Macro liquidity: Fed/PBOC signal, US 10Y, China 10Y, DXY, SHIBOR/DR007, Treasury/TGA or Fed balance-sheet context when relevant
- Risk appetite: VIX, put/call, market breadth, new highs/new lows, major index trend
- Capital flow: US sector ETF flow when available, A-share northbound/southbound if available, margin financing, main-board/STAR/ChiNext turnover structure
- Sector behavior: weekly relative strength vs S&P 500 and CSI 300, volume confirmation, support/resistance, breakout/failure
- Expectations and crowding: earnings revisions, revenue/EPS consensus direction, valuation percentile if available, fund positioning/crowdedness proxies
- Catalysts: earnings calendar, macro data calendar, policy events, product/order/legal/regulatory events

Tag data clearly:

- `【实】` confirmed actual data
- `【预】` market expectation or consensus
- `【估】` model/analyst estimate
- `【待】` reported but not independently verified; avoid using as core evidence

Produce an internal evidence matrix before selecting sectors:

| Market | Constraint | Indicator | Latest value/status | Source | Decision impact |
|---|---|---|---|---|---|
| US | Rates/liquidity/risk | e.g. 10Y, DXY, VIX | value + date | source | supports/weakens/neutral |
| China A | Liquidity/policy/breadth | e.g. DR007, turnover, margin | value + date | source | supports/weakens/neutral |

### 2. Decide Macro Positioning

State whether next week should be `偏进攻`, `中性偏进攻`, `中性偏防守`, or `偏防守`.

Use this chain:

`宏观约束 -> 资金行为 -> 风格偏好 -> 仓位上限 -> 证伪信号`

Do not say "risk appetite improves" without naming the observable variable that improved.

Positioning must include:

- Gross exposure ceiling, expressed as a range, not a false precision point.
- Conditions for increasing exposure.
- Conditions for cutting exposure.
- One correlated-risk warning if several trades depend on the same macro factor.

### 3. Select Bullish and Bearish Industries

Cover both US equities and A-shares. Output at least:

- 2-4 bullish industries across the two markets
- 2-4 bearish industries across the two markets
- At least one bullish and one bearish view for each market unless data quality makes this unsafe; if so, explain why

For each industry, require:

- Industry-level evidence, not only a single stock move
- Market-behavior validation: price, volume, relative strength, breadth, or fund flow
- Catalyst and expectation-gap judgment
- Assumptions and what changes if assumptions fail
- Crowding/overpricing risk check

Avoid hindsight logic:

- Do not mark a current hot theme bullish merely because it is hot.
- Do not mark a falling industry bearish merely because it fell.
- Ask: "what is not yet priced for the coming week?"

Use confidence labels:

- `高`: thesis has macro support, industry confirmation, catalyst, and clear invalidation.
- `中`: thesis has 2-3 supports but one important uncertainty.
- `低`: watchlist only; do not force a trading plan.

### 4. Select Stocks and Trading Plans

For each market, provide a small number of stock calls. Prefer quality over quantity:

- US equities: 2-4 stocks total, including bullish and bearish/avoid candidates
- A-shares: 2-4 stocks total, including bullish and bearish/avoid candidates

Each stock must include:

- Name, ticker/code, market
- Direction: buy/add/hold/reduce/sell/avoid/watch
- Thesis with 2-4 evidence points
- Three-factor score: earnings expectation, valuation/sentiment, liquidity/risk premium
- Expectation-gap quadrant:
  - Good news + high expectation = risk of "priced in"
  - Good news + low expectation = bullish catalyst
  - Bad news + high expectation = bearish shock
  - Bad news + low expectation = possible exhaustion/reversal
- Concrete action plan:
  - Entry trigger or buy zone
  - Staged position sizing
  - Stop-loss or invalidation level
  - Take-profit or reduction plan
  - Hold/sell/buy decision for existing positions
  - Daily tracking indicators and action timetable

If precise price levels are unavailable from reliable data, use technical conditions instead of fabricated numbers, such as "daily close above prior 20-day high with volume above 20-day average".

Trading action rules:

- Do not give a `buy/add/sell` action unless entry, invalidation, sizing, and review date are all present.
- Use `watch/avoid/hold` when data quality is insufficient or the setup lacks confirmation.
- Position sizing must be bounded by thesis confidence and correlated exposure; avoid all-in or certainty language.
- For existing positions, distinguish `持有观察`, `减仓`, and `止损退出` conditions.

### 5. Run TradingAgents-Style Internal Review

Before finalizing, simulate these roles in writing or internally:

- Fundamental analyst: earnings, balance sheet, valuation, guidance
- News/policy analyst: catalysts, official data, regulatory and geopolitical events
- Technical/market analyst: trend, volume, breadth, support/resistance
- Bull researcher: best case and upside path
- Bear researcher: strongest counterargument and downside path
- Trader: translate thesis into action rules
- Risk manager: position sizing, invalidation, drawdown, correlated exposures

Use structured findings rather than long dialogue. The final article should show the result of the debate through assumptions, alternatives, and risk controls.

For each final stock call, include the review result in compact form:

`多头证据 / 空头反驳 / 最终处理 / 证伪信号`

### 6. Validate Before Saving

Check:

- Every major number has a source.
- Dates, units, direction, actual vs expected, intraday vs close are correct.
- US and A-share coverage is explicit.
- Bullish and bearish views are both present.
- Industry calls have industry-level evidence.
- Stock calls include clear trading actions, not vague "关注".
- Each key thesis has assumptions and falsification signals.
- The report includes charts/tables/Mermaid diagrams.
- The article ends with a disclaimer.

If a claim cannot be verified, weaken it, label it, or remove it.

Legal and suitability boundary:

- Write actions as scenario-based public-information plans, not personalized advice.
- Do not imply guaranteed returns, target certainty, or suitability for all investors.
- Keep the disclaimer, but do not use it to justify unsupported precision.

## Output Structure

Save the final Markdown to:

`markdown/finance-weekly-outlook-YYYY-MM-DD.md`

Use the report date, create `markdown/` if needed, and write UTF-8.

Required structure:

```markdown
# 未来一周中美股多空展望|YYYY-MM-DD 至 YYYY-MM-DD

> 执行摘要:...

## 一、结论先行:下周仓位与主线

Include: forecast window, report date, upstream files used, macro positioning, gross exposure ceiling, top bullish/bearish themes.

## 二、数据仪表盘

Include tables and at least one Mermaid diagram showing:
触发变量 -> 数据验证 -> 预期差 -> 交易计划 -> 证伪信号

## 三、行业多空矩阵

## 四、美股个股操盘计划

## 五、A股个股操盘计划

## 六、情景推演:前提、备选观点、证伪信号

## 七、下周关键日历与跟踪清单

## 八、数据来源

## 免责声明
本文仅供参考,不构成投资建议。市场有风险,投资需谨慎。文中观点基于公开信息、特定前提假设和当时可得数据,未来可能因宏观政策、流动性、财报、监管、地缘政治和市场情绪变化而失效。任何买入、卖出、持有或仓位安排都不应被视为个性化投资建议,投资者应结合自身风险承受能力独立决策。
```

## Writing Rules

- Write in Chinese unless the user requests another language.
- Be direct, structured, and source-backed.
- Use "不是A,而是B" only when it exposes a real mechanism mismatch.
- Do not overstate probability; use "基准情景/备选情景/尾部风险" instead of false certainty.
- Give explicit actions for individual stocks, but keep the disclaimer and assumption boundaries clear.
- Prefer concise tables over long prose where decisions need comparison.

讨论

基于 GitHub Discussions。登录 GitHub 即可参与讨论、点赞、订阅更新。