broker-mcp-selector

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs the agent at runtime to "confirm against the live list" hosted at https://github.com/mphinance/awesome-broker-mcp, meaning the agent would fetch that GitHub repo and use its contents to drive broker-selection prompts/decisions.

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

• Direct money access detected (high risk: 1.00). The skill is explicitly about selecting and wiring broker MCP servers for trade execution. It instructs the agent to "actually place the trade" / "submit this order" and to connect an agent to a brokerage, distinguishes servers that "✅ Trades" vs "Draft only" vs "Read-only", names brokers that execute "live on a tool call" (e.g., Robinhood, Tradier, Alpaca, Kraken, etc.), and gives concrete wiring/config examples (e.g. ALPACA_PAPER_TRADE, ALPACA_TOOLSETS). This is specifically designed to enable market orders and live trading execution, so it grants Direct Financial Execution Authority.