PMXT MCP exposes 4 trading tools with complete input schemas and risk annotations. Naming follows verb_noun convention (buildOrder, cancelOrder, createOrder, compareMarketPrices). Descriptions are present and substantive (100-200 chars), explaining WHAT and WHEN. However, parameter descriptions are sparse, many lack guidance on format, constraints, or dependencies. The 'credentials' object is well-documented but exposes sensitive fields as parameters (apiKey, privateKey), violating secret-injection patterns. Output schemas are not documented, LLMs cannot predict response structure. Error handling is absent from visible code. Tool composition is sound (each does one thing), but the credentials parameter design is a critical security flaw.

• Remove credentials from tool parameters. Inject API keys, private keys, and secrets server-side via environment variables or a secure vault. Accept only a 'venue' or 'account_id' parameter to select which pre-configured credentials to use.
• Document output schemas for all tools. Specify what fields each tool returns (e.g., buildOrder returns {signedOrder: string, nonce: number, ...}). This enables tool chaining and LLM planning.
• Add min/max bounds and format descriptions to numeric parameters. E.g., 'amount: number (min 0.001, max 1000000)', 'price: number (0.0 - 1.0 for binary markets)', 'slippage_pct: number (0 - 50, default 2)'.
• Expand compareMarketPrices 'params' into explicit properties. Define what fields the agent should pass (e.g., marketId, outcomeId, side). Use an enum or object schema with required properties.
• Add error handling guidance to tool descriptions. E.g., 'If the order fails with "insufficient balance", call fetchBalance() first. If "market not found", call loadMarkets() to verify the marketId.'