mobiusquant/openmobius-skill

OpenMobius-skill is a plugin for AI coding agents that turns the agent into a chatbot for a particular school of trading analysis called ICT and SMC. It targets four agents: Claude Code, Codex, OpenClaw, and Hermes. Once installed, you can ask the agent trading questions in normal English and it will reach into a built in knowledge base and live market data instead of guessing. The knowledge base is the centerpiece. It holds 964 cards drawn from 130 ICT and SMC teaching videos, split into 380 concept cards and 584 case studies. Each concept card lists identification rules, trading implications, common mistakes, and related ideas. Cards are stored in a local ChromaDB index and retrieved with a small open source embedding model called nomic-embed-text-v1.5, so the lookup runs on your machine and needs no API key. Alongside the cards, the plugin can pull live price data for crypto from Binance, Bybit, OKX, and Hyperliquid, plus China A-shares, Hong Kong stocks, US stocks, and forex. It also exposes more than 60 technical indicators such as RSI and MACD. There are two ways it makes charts: annotate an image the user supplied using PIL, or generate a fresh chart through the lightweight-charts library running inside a headless Chromium controlled by Playwright. The plugin routes each request to one of four workflows called qna, analyze, annotate, and klines. Installation is a single Python script. You clone the repo to a temporary folder and run install.py with a platform flag, for example claude-code or all. The installer copies files into the agent's skills folder, creates a virtual environment, downloads the Chromium build (about 280 MB) and the embedding model (about 274 MB), loads precomputed embeddings, and runs a health check. The first install takes five to ten minutes, updates take under a minute, and each platform install is self contained. The roadmap on the README talks about extending the knowledge base to cover further ICT sub schools, adding Wyckoff, VSA, Volume Profile, and classical price action, turning SMC ideas like Liquidity Sweep and Order Block into computable indicators, and adding chatbot front ends for people who do not use a coding agent. The project is Apache 2.0 licensed and requires Python 3.10 or newer.