Stock Analysis

Stock Analysis

Use this skill to generate an A-share individual stock fundamental analysis report from a stock code. When the prompt asks for 产业链, 卡脖子, 龙头, 买入逻辑, or AI-era suppliers, include the chokepoint lens below instead of producing a generic valuation-only report. If the user asks for 全球股/全球产业链 and says to exclude mainland A-shares, route away from this skill.

Scope

• Source root: ~/.hermes/external-repos/stock-analysis.
• Runtime data collector: stock_full_report.py.
• Data source: AkShare public market/financial endpoints.
• Output convention: write results under the current project's output/ directory unless the user names another destination.
• The report is for research only and is not investment advice.

Safety Rules

• Do not install Python dependencies globally unless the user explicitly asks. Prefer a project-local virtual environment.
• Do not create or edit MCP configs or store API keys unless the user explicitly asks.
• Treat live market/news/financial fetches as external network calls; run them only when the user asks for an actual report.
• If the user only asks for routing or install, do not fetch stock data.

Workflow

1. Confirm the stock code is a six-digit A-share code or infer it only when the name is unambiguous.
2. From a project/work directory, prepare dependencies if needed:

. ~/.hermes/skills/data-science/stock-analysis/.venv/bin/activate

3. Run Phase 1 data collection:

~/.hermes/skills/data-science/stock-analysis/.venv/bin/python ~/.hermes/external-repos/stock-analysis/stock_full_report.py <股票代码>

4. Read output/data_<股票代码>.json.
5. Draft the Markdown report with the Step 0-8 framework from the upstream skill.
6. If HTML is requested, use the upstream shared/ assets and examples/个股研究-中国长城.html as visual reference; write the final HTML under output/.

Chokepoint / 产业链 Mode

Use this mode when the user asks from demand waves, AI产业链, 卡脖子, 龙头, 买入逻辑, 前瞻TAM, qualification cycle, 垄断, 功能性独占, or 价值链向上爬.

Output contract:

• First line: only the A-share leader ticker/name when the user explicitly asks for A-share leaders.
• Do not mix mainland A-share leaders into a global-stock leader answer unless the user explicitly asks for A-shares.
• Put all reasoning, caveats, valuation, and buy logic after the first line.

Frame the thesis from demand backward:

• Demand wave: what secular demand shock is forcing a new architecture or capacity bottleneck?
• Architecture bottleneck: which physical/process/material function becomes scarce?
• Cannot be designed away: why customers cannot easily bypass it through redesign, second source, software substitution, or vertical integration.
• Material revenue: how the company can convert the bottleneck into revenue large enough to matter.
• Qualification cycle: evidence from certification, customer qualification, platform inclusion, long-term orders, capacity reservation, or process/tool approval.
• Early/small preference: prefer smaller or earlier-stage names only after the chokepoint filter passes.

Use this scoring formula as the ranking spine:

excess return =
  major demand trend
  × insufficient supply elasticity
  × low market recognition
  × catalyst
  - valuation / liquidity / dilution / geopolitical risk

Do not let current-period financials dominate when the user's thesis is explicitly qualification-cycle / forward-TAM based. Still name thesis-break conditions: design-out evidence, failed qualification, revenue not scaling after the expected cycle, dilution, customer loss, or policy/geopolitical block.

IBKR/TWS API fallback for US/global watchlists

When the user explicitly requires IBKR/TWS API data (for example port 4002) and the Python ibapi/ib_insync packages are unavailable, use the raw TWS socket protocol rather than silently switching providers:

• Connect to 127.0.0.1:<port> and send the v100 handshake.
• Send START_API (71) before any request.
• Use reqMatchingSymbols (81) to resolve symbols to IBKR contracts/conIds.
• Use delayed market data when subscriptions block live data: REQ_MARKET_DATA_TYPE (59) with type 3.
• Use REQ_MKT_DATA (1) with message version 11 for stock quotes.
• Treat 10167 Requested market data is not subscribed. Displaying delayed market data... as a valid delayed-data status, not a hard failure.

Reference: references/ibkr-tws-raw-api.md. Reusable probe: scripts/ibkr_tws_raw_probe.py.

First Probe

~/.hermes/skills/data-science/stock-analysis/.venv/bin/python ~/.hermes/external-repos/stock-analysis/stock_full_report.py 000001 --max-kline-years 1