trading-brain

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow invokes multiple L1/L3/L5 skills (e.g., news-sentiment, economic-calendar, social-sentiment-scraper, institutional-timeline, mt5-chart-browser) that, by their names and described purposes, can fetch/ingest outsider-authored free text (news/social/calendar/institutional reports) and pass it into the agent/LLM context for sentiment/NLP processing.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill explicitly implements trade execution: it contains a dedicated L6 Execution Engine with components named execution-algo-trading and mt5-integration/mt5-ea-code-generator, explicit code paths for "live" mode that "would invoke MT5 order placement", and returns TradeResult objects with order_id, fill_price, status, slippage, etc. The run() flow moves from risk approval (L5) to execution (L6) and in live mode submits live orders. This is a specific market-order execution capability (placing trades), not a generic API or browser automation, and therefore grants direct financial execution authority.