mcp-server-catalog-reference

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

• Potentially malicious external URL detected (high risk: 0.90). Yes — the skill includes explicit runtime commands that fetch and execute remote code (e.g., "npx -y @modelcontextprotocol/server-postgres", "npx -y @elevenlabs/mcp-server", etc.) and runtime transport endpoints the agent will call (e.g., "https://github-mcp.glitch.me"), meaning external npm packages and HTTP MCP servers are fetched/connected at runtime and can execute code or inject prompts/context.

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

• Direct money access detected (high risk: 1.00). The catalog explicitly lists and provides install/config commands for financial execution servers: payment gateways (Stripe, PayPal, Plaid, QuickBooks), crypto/blockchain providers (Base, Solana, Thirdweb, RPC/private keys), and trading/exchange integrations (Alpaca, CCXT). It shows environment variables for secret keys, warns these servers "handle money" and instructs use of test/sandbox vs live keys. These are specific integrations able to send payments, sign/send on‑chain transactions, and place market orders — i.e., direct financial execution capabilities.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)