A-share Stock Analysis

Overview

Use this skill to manage local A-share watchlists and generate structured trading recommendations through fixed scripts. The standard process does not only look at single-point real-time data; instead, it first checks positions, then judges trends and phases by combining historical and real-time market data, analyzes sectors and the top three strongest stocks within the sector, and finally outputs operational recommendations with process explanations.

Workflow

1. First Determine Request Type

Categorize the task into one of the following types:

Watchlist Maintenance: Add, move, update, view or delete entries in local watchlists and position pools.
Metadata Completion: Batch fill in stock abbreviations or concept sectors.
Data Fetching: Obtain the latest individual stock information, sector leaders or risk events.
Comprehensive Analysis: Output recommendations of Buy, Watch, Hold, Reduce Position or Sell combined with sector benchmarking and risk prompts.

2. Use Fixed Scripts

Prioritize running the scripts built into this skill, do not temporarily rewrite fetching logic:

```
scripts/watchlist_manager.py
```
- Maintain
```
watchlist
```
  and
```
positions
```
- Support
```
--auto-symbol
```
  to reverse-check stock codes based on names
- Support
```
--auto-name
```
  to automatically fill in stock abbreviations
- Support
```
enrich --auto-symbol --auto-name
```
  to batch fill empty code and name fields
- When users correct concept sectors, directly write back to entries in
```
watchlists.json
```
```
scripts/fetch_stock_profile.py
```
- Fetch individual stock information, market snapshots, latest fundamentals, capital flows and concept sectors
```
scripts/fetch_sector_leaders.py
```
- Parse the concept sectors that stocks belong to, and sort out the top three strongest benchmark stocks
```
scripts/fetch_risk_events.py
```
- Prioritize reading the day's risk cache; refresh future performance disclosures, recent financial report releases, share reduction announcements and negative news online when needed
```
scripts/refresh_daily_risks.py
```
- Batch refresh the daily risk cache of stock pools and write to local
```
data/risk_cache/
```
```
scripts/stock_score.py
```
- Summarize data from other scripts, generate fixed output structure and final recommendations

Do not invent another set of data paths unless the scripts are indeed missing required fields. When a script fails, clearly explain the failure point and missing data.

2.1 Fixed Execution Sequence for Comprehensive Analysis

When conducting position analysis, follow the sequence below without skipping steps:

Check current positions
Query historical market data of target stocks
Query real-time market data of target stocks
Judge trends, current phases and whether key abnormal movements occur by combining historical and real-time market data
Query sector data of the target stocks and judge sector trends
Query the top three strongest stocks in the same sector
Analyze the historical trends and today's performance of these three stocks as horizontal references for the target stocks
If financial reports have just been disclosed or there is an abnormal sharp drop, supplement with online searches to determine if there are signals of "falling short of expectations"
Summarize the analysis process and give operational recommendations for each position

2.2 Daily Refresh of Risk Events

Risk prompts should be given as early as possible, rather than only providing explanations after a sharp drop.

Refresh the risk cache of
```
positions
```
at least once per trading day
Risk cache is saved to
```
data/risk_cache/<symbol>.json
```
by default
Prioritize reading the day's cache when analyzing individual stocks; refresh online if the cache is missing
Key points of refresh include:
- Whether there is a performance disclosure plan in the next few days
- Whether annual reports, quarterly reports or performance forecasts have just been released recently
- Whether there are share reduction announcements recently
- Whether there are obvious negative news recently

Recommended commands (paths below are relative to the root directory of this skill, i.e.,

skills/stock-analyzer/

python scripts/refresh_daily_risks.py --bucket positions

python scripts/fetch_risk_events.py 002837 --refresh --pretty

Windows Task Scheduler can directly call
```
run_daily_refresh.bat
```
in the plugin root directory
You can also run
```
register_daily_refresh_task.ps1
```
in the plugin root directory to automatically create scheduled tasks at
```
08:30
```
/
```
12:30
```
every day

Example registration commands:

powershell -NoProfile -ExecutionPolicy Bypass -File .\register_daily_refresh_task.ps1

If a task with the same name already exists:

powershell -NoProfile -ExecutionPolicy Bypass -File .\register_daily_refresh_task.ps1 -Force

The following contents must be explicitly judged in Step 4:

Current trend: Upward, oscillating, weakening or breaking below support
Current phase: Initiation, main uptrend, acceleration, divergence, high-level oscillation, retreat, etc.
Key abnormal movements: Sharp rise, sharp drop, sudden volume increase, volume increase with stagnant price, volume decrease with rebound, breaking below important moving averages, breaking through stage highs

If key abnormal movements occur, this signal must be explained separately, not just giving conclusions.

Auxiliary online searches are allowed and recommended in any of the following situations:

Within 3 trading days before or after the financial report disclosure date
Obvious large drop on the day, e.g.,
```
<= -7%
```
Sharp drop with increased volume
Users explicitly mention "performance falling short of expectations" or "financial report thunder"

The purpose of searching is not to replace market data, but to judge whether there is such additional information:

Whether the financial report falls below market expectations
Whether there are negative statements in performance bulletins, performance forecasts, annual reports or quarterly reports
Whether the mainstream market interpretation attributes the day's sharp drop to financial reports falling short of expectations

3. Prioritize Local Coverage for Sectors

First check the

concept_sector

field in entries of

watchlists.json

. If it is not filled, then parse the concept sector in real-time via

fetch_stock_profile.py

fetch_sector_leaders.py

. When users explicitly correct concept sectors, directly write back to the entries via

watchlist_manager.py

3.1 Fill in Names First, Then Sectors

If users only provide stock names without codes, first use

watchlist_manager.py

to automatically reverse-check codes:

Single entry addition:
```
add <name> --auto-symbol
```
Batch completion:
```
enrich --auto-symbol
```

If the

name

field of local entries is empty, prioritize using

watchlist_manager.py

to automatically fill in stock abbreviations:

Single entry addition or update:
```
add --auto-name
```
or
```
update --auto-name
```
Batch completion:
```
enrich --auto-name
```

If real-time completion of codes, names or sectors fails:

Clearly explain the source of failure, such as
```
quote_snapshot
```
or
```
stock_info
```
Retain existing local data, do not overwrite with guessed values

4. Output in Fixed Structure

When conducting comprehensive analysis, always output in the following order:

Basic information of the target stock
Historical market data and trend judgment
Real-time market data and key abnormal movements
Belonging concept sector and sector trend
Benchmark table of the top three strongest stocks in the sector
Reference of historical trends and today's performance of benchmark stocks
Scores of technical aspects, fundamental aspects and risk aspects
Recommended action:
```
Buy
```
,
```
Watch
```
,
```
Hold
```
,
```
Reduce Position
```
,
```
Sell
```
Position recommendation
Stop-loss and take-profit conditions
Risk event prompts
Brief conclusion

Use the template in

references/对标分析模板.md

5. Conservatively Handle Failure Scenarios

If real-time data is incomplete or unavailable:

Explain which script failed and which fields are missing.
If users have provided data, it can be used, but must be clearly marked as user-provided.
Do not give strong conclusions when sector benchmarking or risk events are unavailable.
If
```
concept_sector
```
fails to be parsed, must clearly prompt users whether to manually supplement sector information and re-run the analysis, cannot silently skip the sector link.

Data Files

This skill stores local status under

data/

```
data/watchlists.json
```
: Watchlists and position pools
```
data/risk_cache/
```
: Daily risk scan results saved by stock

See

references/股票池结构.md

for JSON structure.

Reference Materials

Load these files as needed:

```
references/数据来源.md
```
- Data source priorities, real-time data assumptions, and fallback methods for failures
```
references/股票池结构.md
```
- Local JSON layout and field descriptions
```
references/A股评分规则.md
```
- Scoring rules for technical aspects, fundamental aspects and risk aspects
```
references/对标分析模板.md
```
- Fixed benchmark table and conclusion structure

stock-analyzer

NPX Install

Tags

SKILL.md Content (Chinese)