Agent wallet CLI — a command-line interface to the Finance District MCP wallet server
Critical gaps across all dimensions. Tool definitions lack proper schemas, descriptions are minimal or missing, and parameter documentation is sparse. 'wallet' and 'prism' tools have only a 'method' parameter with no enum or validation, making them opaque proxies rather than well-defined tools. No output schemas documented. Error handling is absent from tool definitions. Only 'register', 'login', 'verify' have basic descriptions; others are generic or incomplete. This server exposes MCP as a thin CLI wrapper without proper tool abstraction.
Show or manage persistent configuration (~/.fdx/config.json)
Sign in to an existing account (sends email OTP)
Remove stored credentials
Prism MCP tools (payments, settlements, accounts)
Register a new account (sends email OTP)
List available services
Check authentication status
wallet and prism tools are opaque method dispatchers with no schema. Single 'method' parameter accepts arbitrary strings with no enum, validation, or documentation of available methods. LLM cannot determine valid options or what each method does.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields. Responses are unstructured and unpredictable.
status, logout, services have minimal descriptions (under 30 chars). Insufficient context for LLM tool selection. 'Check authentication status' and 'Remove stored credentials' lack detail on when to use them or what they return.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | <=2025-11-25 | v2 |
Complete registration or login by submitting the OTP code
Wallet MCP tools (DeFi, transfers, X402 payments)
config tool has three optional parameters (action, key, value) with no clear documentation of valid action values, required combinations, or behavior when omitted. Parameter relationships undocumented.
No error handling guidance in tool definitions. No recovery hints, categorization of retryable vs fatal errors, or actionable error messages. Agents cannot self-correct on failure.