Goblin is a gateway/proxy server that exposes meta-tools for discovering and invoking other MCP servers. The 4 tools (describe_tool, invoke_tool, describe_server, search_servers) are well-named with action verbs and have clear, detailed descriptions (120-180 chars each). All tools have proper input schemas with type definitions and parameter descriptions. However, output schemas are not explicitly documented in the visible source code, and error handling guidance is minimal. The tools follow good composition patterns (each does one thing), but lack concrete field-level return documentation that would allow agents to plan downstream calls. Security is addressed (invoke_tool validates arguments), but permission gates and audit logging are not evident in the provided code snippets.

• Document the exact output structure for each tool. E.g., describe_tool should explicitly state it returns {name, description, inputSchema, riskLevel}. invoke_tool should document {success, result?, error?, retryable?}. This allows agents to plan multi-step sequences.
• Add error classification guidance to invoke_tool. E.g., 'If tool not found, error will include available_tools array. If validation fails, error includes invalid_field and constraint_violated. Both are user-fixable; retry after correction. If server unreachable, error is retryable.'
• Add a limit parameter to describe_server (default 100, max 1000) and search_servers (default 50, max 500) to enforce result bounds and prevent context window exhaustion.
• Document required vs optional chaining fields in responses. E.g., after invoke_tool succeeds, what IDs (server_id, tool_name, etc.) does the result include so agents can compose follow-up calls?
• Define the semantics of search_servers with an empty query parameter. Either make it required, or explicitly state 'empty query returns all servers sorted by name'.