stock-trading-analyst

これはレビュー時に保存された抜粋です。完全で最新の内容は外部ソースを確認してください。

---
name: stock-trading-analyst
description: Use when the user needs an A-share stock watchlist, short-term theme rotation review, or trading-plan risk audit based on user-provided market data. Produces evidence-based sector analysis, signal confidence, risk warnings, and a non-advisory action checklist for retail investors, research assistants, and trading educators.
version: 0.1.0
license: MIT
tags: [finance, stock, trading, a-share, risk-control]
author: SkillLens Demo Team
---

# stock-trading-analyst

## When to use

Use this skill when the user provides A-share market data and asks for:

- 今日题材轮动、涨停归因、资金生态分析
- 短线候选股筛选、观察池复盘、交易计划风控检查
- 个股异动原因拆解、板块强弱比较、情绪周期研判
- "帮我看看这批股票哪个更值得观察"

**Not suitable for**:直接给出买入 / 卖出指令、承诺收益、代客理财、荐股收费、绕过投顾合规要求、处理内幕信息或未授权账户数据。

## Target users

- 有基础交易经验、需要结构化复盘的 A 股短线投资者
- 金融自媒体 / 投教团队，用于把盘面数据转成可解释的复盘框架
- 证券研究助理，用于整理公开市场数据、生成观察清单初稿
- 量化或半自动交易团队，用于人工复核前的信号解释层

Estimated frequency: daily after market close, plus intraday review during high-volatility sessions.

## Value proposition

Most generic LLM stock prompts jump directly to "看好 / 不看好" and ignore evidence quality, position risk, data freshness, and compliance boundaries. This skill focuses on **decision support, not trading instruction**:

- separates market facts, inferred signals, assumptions, and risks;
- scores signal confidence instead of pretending certainty;
- keeps high-risk outputs behind a human-review checklist;
- uses schema validation so reports can be compared across days;
- explicitly refuses return promises and direct personalized investment advice.

Expected value: reduce a 60-90 minute manual replay of limit-up themes, capital flow, turnover, and risk notes into a 15-25 minute structured review, while preserving human judgment.

## Inputs

| Field | Type | Required | Notes |
|---|---|---:|---|
| `market_date` | string | yes | Trading date, e.g. `2026-05-07` |
| `universe` | enum | yes | `a_share`, `hk_stock`, `us_stock`; MVP tuned for `a_share` |
| `rows` | array | yes | User-provided table rows; one row per stock or concept |
| `scenario` | enum | optional | `theme_rotation`, `watchlist`, `risk_audit`, `education` |
| `risk_profile` | enum | optional | `conservative`, `balanced`, `aggressive`; defaults to `balanced` |
| `holding_context` | object | optional | Positions, cost basis, or target watchlist; do not include account credentials |

Recommended row fields:

```json
{
  "code": "000001",
  "name": "示例股份",
  "concepts": ["AI应用", "金融科技"],
  "price_change_pct": 7.2,
  "turnover_rate": 18.5,
  "volume_ratio": 2.6,
  "net_inflow_cny": 125000000,
  "limit_up_reason": "题材催化 + 资金回流",
  "news": ["公告摘要或公开新闻链接"],
  "data_source": "user_upload",
  "timestamp": "2026-05-07T15:10:00+08:00"
}
```

## Workflow

1. **Validate input**: check required fields, timestamp freshness, duplicated codes, abnormal values, and missing data source.
2. **Classify scenario**: infer whether the user wants theme rotation, watchlist screening, risk audit, or education; ask a follow-up question if ambiguous.
3. **Normalize evidence**: separate objective fields (涨幅、换手、量比、资金净额), user notes, public news, and model inference.
4. **Theme rotation analysis**: rank concepts by breadth, limit-up strength, capital inflow, turnover sustainability, and intraday consistency.
5. **Candidate scoring**: score each candidate on signal strength, liquidity, catalyst clarity, crowding risk, and data confidence.
6. **Risk and compliance gate**: detect direct advisory wording, overconfident return claims, high volatility, illiquidity, ST / delisting risk, single-source evidence, and missing human review.
7. **Generate report**: output JSON following `assets/stock_signal.schema.json`, plus a concise markdown summary for humans.
8. **Validate output**: run `scripts/validate_signal.py` against the JSON schema. If validation fails twice, return `_schema_failed: true` and list the validation errors.

## Analysis framework

### Scenario fit

The skill must state whether it is do