枢门财经 | ShumenFinance

This skill is a deterministic mainland China A-share wrapper around the Tianshan APIs. It always calls https://tianshan-api.kungfu-trader.com. It runs local .mjs flows with node. At runtime it only reads bundled prompt/reference assets from this skill repo and does not write local files. The outbound network surface is intentionally fixed to that single base URL and is not user-configurable. It does not shell out to subprocesses, install dependencies, or fetch additional code at runtime. It requires host-injected runtime authentication and platform metadata. Auth material is read only from runtime environment variables and is never accepted from user prompts.

Current rollout note:

• this skill requires host-provided runtime auth
• current public build only exposes products whose backend routes have been revalidated against the current production contract
• unresolved products stay behind the local flag KUNGFU_ENABLE_EXPERIMENTAL_PRODUCTS=1

Latency-sensitive usage note:

• when the host message already明确指向 基本面分析，优先直接调用 finance_basic_indicators
• when the host message already明确指向 技术面分析，优先直接调用 price_levels
• do not spend extra tool calls on filesystem probing, environment echo, or duplicate skill discovery
• assume runtime auth is already injected; if auth is missing, fail fast instead of probing
• if a tool call returns {"success":false,"error":{"code":"AUTH_INVALID"...}} or PERMISSION_DENIED, stop immediately and tell the user the protected capability is unavailable under the current runtime auth
• on AUTH_INVALID or PERMISSION_DENIED, do not retry, do not inspect environment variables, and do not read source files for debugging

Host-agent async note:

• if the host agent is serving any OpenClaw external channel and the user request clearly implies deep multi-step research, do not expand this skill into a long synchronous interaction chain
• long-form research should be queued as an async background job by the host agent, while this skill remains focused on deterministic data collection
• keep /shumen_finance XXX 基本面分析 and /shumen_finance XXX 技术面分析 on the fast path unless the user explicitly asks for a deep comprehensive report

Treat the current product list as a skill-side contract under migration. Only the rollout-enabled products below should be considered stable and public for the current release build.

Do not use this skill with secrets, personal data, or other sensitive user content. User requests and derived analysis context may be transmitted to the upstream Tianshan API.

Treat this file as the high-level routing guide. Do not treat it as an API catalog.

Scope

• Mainland China A-shares only
• Single stock or single sector only
• No whole-market screening
• Not for US stocks
• Not for Hong Kong stocks
• Not for cryptocurrencies
• Not for futures or forex
• No free-form SQL
• No user-facing base URL
• No user-facing runtime credential parameters

Public Intents

Route the user into the currently enabled intent families below. Do not expose internal product names unless needed for implementation.

1. Stock Snapshot

Use when the user wants one A-share stock's current state, recent走势, K-line window, holders,筹码分布,支撑压力位, or other narrow deterministic data.

Default behavior:

• Prefer the narrowest data product that answers the question
• If the stock code is unknown, prefer instrument_name
• If code mode is used, exchange_id must be SSE or SZE

2. Finance Analysis

Use when the user wants one A-share stock's基础财务指标或基础财务上下文。

Default behavior:

• Prefer finance_basic_indicators for narrow factual questions
• Prefer finance_context for lightweight finance context

3. Sector Analysis

Use when the user asks about one A-share industry or sector.

Default behavior:

• If the user gives a standard sector name, sector detail products may use sector_name directly
• The current release build does not expose fuzzy sector resolution
• Treat sector_id as an internal identifier, not a user-facing requirement

Not Yet Enabled In Experience Build

The following capabilities are still under route-contract cleanup and should not be used in the current public rollout:

• full finance-analysis bundles
• unusual movement analysis bundle
• money flow
• fuzzy sector resolution / similar sectors
• market news lookup

Intent Routing

Pick one default route from the list below.

• User asks "这只股票现在怎么样 / 最近走势 / 看看价格": Route to Stock Snapshot
• User asks "分析财报 / 财务怎么样 / 估值怎么看": Route to Finance Analysis
• User asks "分析白酒 / 算力 / 某个板块 / 某个概念": Route to Sector Analysis

If the user asks for whole-market screening, movement-analysis bundles, or market-news bundles, do not use this skill in the current release build.

Input Rules

Stock Identification

For single-stock requests, use exactly one:

• --instrument-name
• --instrument-id and --exchange-id

Never mix both modes. If the stock code is unknown, prefer --instrument-name. When using code mode, --exchange-id must be SSE or SZE.

Stock Date Mode

For stock products and flows, use:

• --target-date
• optional --visual-days-len

Sector Mode

• For resolve_sector and similar_sectors, use --query
• For sector detail products, use exactly one:
  • --sector-name
  • --sector-id

Prefer --sector-name. Only reuse --sector-id after the backend has already returned it.

Hidden Parameters

Do not generate or expose these from the model side:

• start_date
• end_date
• is_realtime
• limit
• days
• periods

Internal Building Blocks

The internal deterministic product contract lives here:

• data_products.md

Use that file only when you need exact internal product names, allowed parameter shapes, or output keys. Do not dump that whole catalog into the user-facing reasoning path by default.

Research Methodologies

Documented research methods live here:

• README.md
• stock-analysis
• sector-analysis

Use them as methodology references for deep research. They are not public CLI entrypoints yet.

Prompt assets are flow-specific. Do not mix finance-analysis formatting requirements into movement-analysis outputs, or vice versa.

Standard Commands

Stock Snapshot

node scripts/flows/run_data_request.mjs --product price_snapshot --instrument-name 贵州茅台 --target-date 20260301

Sector Detail

node scripts/flows/run_data_request.mjs --product sector_performance --sector-name 白酒 --target-date 20260301