English | 中文

Digital Oracle 📈

Digital Oracle is an open-source Skill that lets AI Agents mine macro-event trends from massive financial data.

Works with OpenClaw / Claude Code / Cursor / Codex.

We live in an era of extreme noise. Social media is flooded with emotional predictions — someone says housing is about to crash, someone says gold is going to the moon, someone says war breaks out tomorrow. These opinions chase crowd sentiment rather than rational analysis grounded in objective data.

But trading data is different.

Price is absolutely rational — when someone puts real money on an outcome, they think a lot harder than when they post a short video.

This is the core insight of the Efficient Market Hypothesis: all public information is already priced in. Everything is in the chart.

Digital Oracle turns this insight into an executable tool. It plugs into 13 authoritative financial data sources — from prediction markets like Polymarket and Kalshi, to US Treasury yield curves, CFTC institutional positioning, SEC insider trades, central bank rates, crypto derivatives, and China A-share order-size fund flow.

It doesn't read newspapers, news articles, short videos, or podcasts. It answers questions about housing prices, gold trends, Bitcoin cycles, and military conflict probabilities purely through price signals mined from financial data — delivering structured probability estimates with full reasoning chains.

In a sense, it's a digital oracle for the new era.

What can it answer?

• "What's the probability of WW3?"
• "Will there be a US recession this year?"
• "Is AI in a bubble?"
• "Is now a good time to buy gold?"
• "Has Bitcoin bottomed?"
• "Is NVDA options premium overpriced?"

If there's a market pricing an outcome, Digital Oracle can give you a probability estimate backed by trading data.

Data Sources

Provider	Data Type	Purpose
Polymarket	Prediction market contracts	Event probability pricing
Kalshi	SEC-regulated binary contracts	US political/economic events
Stooq	Stocks/ETFs/FX/Commodities	Price history and trends
Deribit	Crypto derivatives	Futures term structure, options IV
US Treasury	Treasury yields	Yield curves, inflation expectations
CFTC COT	Futures positioning	Institutional direction (smart money)
CoinGecko	Crypto spot	BTC/ETH price, market cap
SEC EDGAR	Insider trades	Form 4 buy/sell signals
BIS	Central bank data	Policy rates, credit-to-GDP gaps
World Bank	Development indicators	GDP, population, trade
Yahoo Finance	US options chains	IV, Greeks, put/call ratio
Eastmoney	China A-share	Quotes, OHLCV, order-size fund flow, sector rotation
Web Search	Web search	VIX, CDS, and other supplementary data

All APIs are free and require no API keys.

Installation

OpenClaw

clawhub install digital-oracle

Other AI Agents (Claude Code / Cursor / Codex / ...)

Just tell your agent:

Install this open-source project and read SKILL.md as your working instructions: https://github.com/komako-workshop/digital-oracle

The agent will clone the repo, read the methodology, and call the providers on its own.

Prerequisites

• uv — Python package manager, used to run skill scripts at runtime
• 11 out of 12 data sources have zero external dependencies (pure Python stdlib). Options chain analysis requires an extra install:

uv pip install yfinance

How It Works

1. Understand the question — decompose into core variables, time window, and priceability
2. Select signals — pick 3+ independent data sources based on question type
3. Fetch in parallel — use gather() to call multiple providers concurrently
4. Contradiction analysis — find disagreements between markets, explain why both can be right
5. Output report — structured multi-layer signal tables + probability estimates + scenario analysis

Project Structure

digital-oracle/
├── SKILL.md                # Skill definition (read by OpenClaw)
├── digital_oracle/         # Python source code
│   ├── concurrent.py       # Parallel execution utilities
│   ├── http.py             # HTTP client abstraction
│   ├── snapshots.py        # HTTP response recording/replay (for tests)
│   └── providers/          # 12 data providers
├── references/             # API reference
│   ├── providers.md        # Provider API docs
│   └── symbols.md          # Trading symbol directory
├── scripts/                # Demo scripts
└── tests/                  # Unit tests + fixtures

Design Principles

• Zero dependencies first — 12/13 providers use only the Python standard library, no pip install needed
• Dependency injection — all providers accept an optional http_client parameter for easy testing
• Partial failure tolerance — one data source going down doesn't break the rest
• Snapshot testing — record real HTTP responses, run tests offline in CI

License

MIT © 2026 komako-workshop — see LICENSE.