fincept-debug

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly instructs the agent to consult and interpret public third‑party sources—e.g., "Check provider status page" and "Check provider API documentation" in the WebSocket Debugging Workflows and to capture/inspect raw messages from external providers—actions that ingest untrusted external content and can change debugging/reconnection behavior.

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

• Direct money access detected (high risk: 1.00). The skill explicitly covers trading/order execution and broker integrations: it mentions "trading systems (order execution, paper/live mode, broker APIs)", contains workflows for "Order Execution Failures", code examples and commands that validate and route orders, and a tauri command example place_order which dispatches to paper_trading::place_order or broker::place_order. It names broker-specific adapters (Alpaca, Binance, broker_*.rs) and symbol/parameter translation for actual broker APIs. These are specific, explicit capabilities for placing market/broker orders (i.e., moving money/positions), not generic debugging tools.