alpaca-trading-backtest

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

• Third-party content exposure detected (high risk: 0.75). The required workflow fetches historical market data via the Alpaca CLI (e.g., alpaca data bars/quotes/... --quiet > raw/*.json), and those CLI responses are treated as readable JSON/prose that the agent then parses and includes in run artifacts/LLM context; this is outsider-authored content from a third-party public data provider (Alpaca) rather than user-authored text.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill includes a runtime installation command that fetches and builds remote code — "go install github.com/alpacahq/cli/cmd/alpaca@latest" — which executes externally-sourced code during runtime and the workflow depends on the resulting CLI, so this is a runtime external dependency that can execute code.

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

• Direct money access detected (high risk: 1.00). The skill is explicitly for trading/backtesting using the Alpaca CLI (a brokerage API). It requires Alpaca authentication (API keys/env vars), instructs use of the Alpaca CLI for data and profiles, and even describes an optional paper forward-validation handoff that includes an alpaca_order_adapter.py and strategy_runtime for automated (paper) orders. Although it warns not to submit live orders during a backtest, the documented tooling and authentication are specific to a brokerage API capable of placing market orders. This is a purpose-built financial trading tool rather than a generic interface, so it grants direct financial execution capability (via Alpaca).