同花顺 iFinD 接入 Skillv0.5.1

A financial analyst asks for '贵州茅台近三年营收和毛利率' in plain Chinese and the skill routes the question through iFinD's smart_stock_picking and returns quotes, financials, and reports without manually writing any API endpoint. All data is restricted to the authenticated iFinD account, so the answer stays compliant with licensed market data use. The skill resolves formal Chinese stock names directly and uses a small alias table for nicknames like '茅台' or '宁王'.

A trader or portfolio manager queries 涨停数据, 龙虎榜, 两融, 北向持股, 解禁 and 停复牌 across the A-share market. The skill maps colloquial event questions to iFinD's formal endpoints such as /smart_stock_picking and /date_sequence, returning structured JSON that can be embedded in a dashboard. Cache tokens are reused automatically so repeated daily queries do not disturb the user or leak credentials.

An OpenClaw, Hermes, or Codex agent is installed for a user who already has iFinD credentials but wants LLM-native access. The installer runs scripts/install_skill.sh into ~/.openclaw/workspace/skills/tonghuashun-ifind-skill and then walks the user through copying a refresh_token from the iFinD account details page. The skill exchanges it for an access_token and never asks for the user's username or password.

A regulated firm needs a chatbot that answers only from the licensed iFinD source and explicitly refuses to fill gaps with Tencent Finance, Eastmoney, or other public websites. When iFinD returns an error, permission denial, or ambiguous security name, the skill asks the user for a more precise code or name instead of switching data sources. This keeps the assistant aligned with data vendor licensing and internal compliance rules.

A power user needs a less common iFinD interface not covered by smart-query. The skill first attempts natural-language routing, then endpoint-list and endpoint-call, and only then falls back to api-call using local routing.md and use-cases.md references. The payload is passed as native iFinD JSON without secondary rewriting, preserving exact query semantics.

The core natural-language query skill is distributed freely on ClawHub and GitHub, but every query requires the user's own authenticated iFinD access token. Revenue is indirect: the vendor acquires paying iFinD subscribers, while the skill drives stickiness and data consumption. Up-sell paths include premium routing references and enterprise priority support.

Financial institutions install the skill inside their own OpenClaw or Hermes environment and connect it to their existing iFinD enterprise account. The skill never handles credentials in plaintext, only caching double tokens locally and going through official /get_access_token renewal. Pricing can be per-seat license or annual site license bundled with integration services.

Users can enable IFIND_ROUTE_LLM_ENABLED to enhance complex natural-language parsing through an OpenAI-compatible endpoint, using their own OPENAI_API_KEY. The default remains a deterministic local router to avoid extra cost. The vendor can monetize a managed routing service or bundled model credits for teams that want turnkey accuracy.