What does this skill do, and when should you use it?
global-stock-data is an Agent Skill by Simon Lin distributed as one self-contained SKILL.md with embedded Python, consolidating market data from 11 sources — Sina, Tencent, Eastmoney, Yahoo, SEC EDGAR, FINRA, CBOE, Treasury, CFTC and Nasdaq — into tools an AI assistant can call directly. The architecture spans 13 layers and 30+ endpoints, requires only the `requests` library, and needs no API keys. V2.0 is official-source-first: it adds CBOE's official options feed (full Greeks + IV + 0DTE + unusual-flow detection), FINRA market-wide daily short volume (12,112 symbols), the SEC EDGAR filing stream, and a free market-wide XBRL screener. Every source is labeled with an S/B/C compliance tier with terms quoted verbatim, and the project ships code only — never market data; the author even removed a working HKEX CCASS layer that violated HKEX's terms.
- Fetches US/HK real-time quotes (Sina 36/25 fields, Tencent 71/78 fields, Eastmoney push2 unified secid interface)
- Pulls daily/weekly/monthly/minute K-lines (US data back to 1984) and computes MA/EMA/MACD/RSI/KDJ/Bollinger locally in pure Python
- Reads three financial statements, key metrics (US 49 / HK 75 fields), analyst estimates, institutional holdings, and 503 GAAP metrics from SEC XBRL
- Queries CBOE official option chains (delta/gamma/vega/theta/rho + IV + 0DTE filter + unusual-flow detection) and FINRA market-wide daily short volume
- Searches EDGAR same-day filings (Form 4 / 8-K / 13F), full-text search back to 2001, and runs market-wide XBRL cross-sections via EDGAR frames
- Adds macro data: Treasury yield curve, CFTC COT reports, Nasdaq earnings calendar
- Retail investors asking in natural language: have the assistant report AAPL's price, PE/PB/ROE, target price and institutional holders
- Options traders tracking 0DTE: view NVDA's same-day expiry chain with Greeks and unusual flow (vol/OI > 1)
- Short-interest analysis: check TSLA's short-volume ratio trend this week from FINRA Reg SHO
- Quant researchers batching data: use EDGAR frames to pull one XBRL tag across the whole market (e.g. CY2025Q1 net income for 5,309 companies) as a free screener
- Compliance/event monitoring: get today's Form 4 insider filings for a watchlist, or full-text search which 8-Ks first mentioned a keyword
- HK investors checking Tencent (00700) quotes, K-lines and Chinese-language statement line items
- Users who need to commercialize or redistribute tier-B/C source data (CBOE options, Yahoo, Eastmoney) — the project mandates tier-S sources only for commercial use, and CBOE requires advance approval plus a license agreement
- Users needing HK options or CCASS/southbound holdings data — the options layer is US-only, and HKEX scraping code was deliberately removed because its terms prohibit it
- Traditional users without an AI coding assistant who won't copy Python — this is a SKILL.md with embedded code, not a finished app or hosted API
How do you install this skill?
- You must replace SEC_CONTACT with your real name and email before using SEC layers, or SEC will reject requests.
- C-tier sources (CBOE/Yahoo/Eastmoney/Sina/Tencent) are for personal research only; for commercial use rely solely on S-tier sources and verify FINRA (B-tier) terms yourself.
- SEC/FINRA/CBOE/Yahoo may be unreachable or unstable from mainland-China networks; use a proxy or prefer Eastmoney/Sina/Tencent sources.
- This static review did not execute any endpoint; upstream API changes can silently break functions — smoke-test before relying on them.
- Compliance tier conclusions date from a 2027-07-24 terms reading; re-verify source terms before commercial use.
- Shell / CLI
- Network access
- Local filesystem
Python 3.9+requests
For Claude Code, follow the README's three-step quick start:
mkdir -p ~/.claude/skills/global-stock-data
curl -o ~/.claude/skills/global-stock-data/SKILL.md \
https://raw.githubusercontent.com/simonlin1212/global-stock-data/main/SKILL.md
pip install requestsFor Codex / OpenClaw and other assistants, the README says to paste SKILL.md into your system prompt or project context file — no separate install command is documented.
Note: before using any SEC layer, set SEC_CONTACT in SKILL.md to your real name and email, or the code will raise an explicit error.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Show NVDA's 0DTE options flow today — any unusual activity?
- What's TSLA's short volume ratio trend this week?
- Any Form 4 insider filings today for my watchlist?
- Does AAPL's MACD show a golden cross? Also pull the latest income statement
After installing, launch Claude Code and talk about US/HK stocks in either language — the skill activates automatically (SKILL.md's When to Activate section lists trigger keywords and scenarios covering quotes, K-lines, filings, options, fund flow and technicals). Core flow: the assistant locates the right layer via the endpoint routing table, then runs the embedded Python (only dependency: requests). Worth knowing: Yahoo cookie+crumb is auto-fetched and cached; all V2.0 layers route through a unified official_get() with a thread-safe per-source rate limiter (SEC 8 req/s, FINRA/CBOE 4 req/s); exceptions distinguish DataNotAvailable (safe to fall back to the next date) from RuntimeError (config/throttle errors that must surface); Layers 6.1/9/10/11 are US-only and raise a clear error for HK tickers. From mainland-China servers, Yahoo/SEC may be flaky — fall back to Eastmoney/Sina/Tencent sources or use a proxy.
What are this skill's strengths and limitations?
- Zero auth, zero API keys, single dependency on requests — very low setup cost
- CBOE Greeks/0DTE, FINRA short volume, EDGAR full-text search and the market-wide screener are capabilities yfinance and most open-source options lack
- Serious compliance work: terms read and quoted verbatim per source, the HKEX-violating CCASS layer removed, commercial boundaries clearly stated
- Solid engineering: SEC User-Agent validation, positive detection of S3 403 AccessDenied, a Tencent field-offset table calibrated against live responses with reproduction commands
- Tier-C sources (Yahoo/Eastmoney/Sina/Tencent/CBOE) are personal-research only; commercial use is restricted or needs licensing
- Nasdaq's earnings-calendar terms are marked unverified; Treasury/CFTC tiers are assigned by convention rather than term-by-term review
- No standalone test suite; unofficial front-end endpoints mean field offsets can break when sites change (the Tencent table was already revised via an issue)
- HK K-lines now have a single available source (Yahoo) after Sina's HK K-line endpoint died — a single point of failure
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Global Stock Data — Full-Stack US & HK Market Data Skill this page | 62 · Recommended | ★ 1.7k | 13d ago | Apache-2.0 |
| TradingView MCP AI Trading Intelligence | 50 · Use with care | ★ 5.1k | 11d ago | MIT |
| Claw-Screener Stock Screener | 51 · Use with care | ★ 10 | 4mo ago | MIT |
| US Stock Market Sentiment Monitor | 54 · Use with care | ★ 1.1k | 6mo ago | MIT |
| Personal Finance Skill (AI-Powered Personal CFO) | 50 · Use with care | ★ 25 | 8mo ago | MIT |
The README explicitly names [a-stock-data](https://github.com/simonlin1212/a-stock-data) as a sister project: it covers China A-shares while global-stock-data covers US and HK markets; installing both causes no conflicts.
How did FollowSkills review this skill?
Strong safety practices: mandatory SEC UA declaration with fail-fast, thread-safe rate limiting (8 req/s vs SEC's 10), per-source compliance tiers with verbatim terms, proactively excluded the HKEX scraper layer, explicit 'code not data' distribution. Deductions: still automates access to C-tier sources (Sina/Tencent/Eastmoney/Yahoo front-end endpoints), a compliance gray zone; data flow depends on multiple unsigned third-party endpoints with no isolation from external behavior changes.
Self-contained code, clean exception semantics (DataNotAvailable/RuntimeError/ValueError), content-driven 403 disambiguation of S3 AccessDenied, locked rate limiter. But no committed test suite or CI covering key paths is visible in the sources; several layer implementations (CBOE, Layers 9–12) are truncated in the provided SKILL.md, so completeness of all claimed endpoints cannot be confirmed; many 'tested' claims are not independently reproducible in a static read.
Clear trigger conditions (activation list + keyword table), explicit non-fit boundaries (US-only layers guard against HK tickers via assert_us_ticker, HK options declared unsupported), good Chinese support. Deduction: core new layers depend on overseas services (SEC/FINRA/CBOE/Yahoo); the README itself acknowledges mainland-China direct connections may be unstable and recommends proxies.
Well-layered docs, routing quick-reference table, semantic versioning v2.0.3, detailed CHANGELOG, Apache-2.0 license, community contribution credits. Deductions: no CI/test directory evidence, individual maintainer, tier conclusions are time-sensitive with no stated re-verification cadence, and the Nasdaq earnings calendar is offered despite unverified terms.
Claims 30+ endpoints across 11 sources, zero-auth, minimal dependency (requests only), with differentiated value (CBOE Greeks, FINRA short volume, EDGAR screener) over yfinance-style alternatives. However, static review cannot confirm outputs are directly usable; v2.0.2 records that two quote functions previously crashed 100% of the time, showing real endpoint fragility — external API changes can silently break features.
CHANGELOG includes before/after fix tables, internal consistency checks (PE = price/EPS, market cap cross-checked against share count), self-reproduction curl commands, and third-party corroboration via community issues. But all 'verified' claims are author self-reported; no independent test suite or CI workflow in the repo to cross-check, so reproducibility cannot be confirmed statically.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →