Doberman Core is a security/authorization framework, not a production MCP server. The 6 tools are defined in a test fixture (fake_tool_server.py), not in actual server code. Tool definitions are minimal: descriptions are present but generic (10-30 chars), input schemas lack parameter descriptions entirely, and no output schemas are documented. The tools themselves (fs_write, fs_delete, shell_exec, net_get, send_email, send_message) are high-risk operations with no error handling guidance, permission gates, or recovery instructions. This is a reference implementation for testing, not a production-grade MCP server.

• Move tool definitions from test fixtures into actual MCP server code with explicit tool registration via the MCP SDK.
• Add parameter descriptions to all input schemas. E.g., 'path: The file system path (absolute or relative) to write to' and 'content: The text or binary content to write (will overwrite if file exists)'.
• Document output schemas for each tool. E.g., fs_write returns {success: bool, path: string, bytes_written: int, error?: string}.
• Expand tool descriptions to 50-200 chars with context. E.g., 'Write content to a file at the specified path. Overwrites existing files. Returns byte count written or error details.'
• Add permission gates and audit logging for destructive tools (fs_delete, shell_exec, send_email). Require explicit authorization checks before execution.