This server defines 3 tools with critically underdeveloped schemas and descriptions. All three tools are inferred from test fixtures (tests/integration/mcp_server_test.ts) rather than directly visible in server implementation code. The tool definitions lack proper parameter type information, descriptions are minimal, and there is no evidence of actual tool implementation. The 'agent_group-*' naming violates basic naming conventions, tools should use verb_noun patterns (e.g., execute_file_operation, query_network, execute_database_query) rather than generic group names. Input schemas show only 'prompt' and optional 'context' as generic object types, with no constraints, enums, or actionable guidance. Output schemas are completely undocumented. Error handling and recovery guidance are absent.

• Rename tools to use verb_noun pattern with specific action names: 'execute_file_operation', 'query_network_resource', 'execute_database_query' instead of 'agent_group-1/2/3'.
• Expand descriptions to 50-200 characters, stating WHAT the tool does, WHEN to use it, and WHAT it returns. Example: 'Execute file system operations (read, write, list, delete, move, copy). Use when the task requires filesystem access. Returns operation status and results.'
• Document the output schema for each tool, specifying returned fields, types, and what chaining IDs are included for downstream calls.
• Add explicit parameter descriptions and constraints. For 'prompt': describe expected format, length, and example intent. For 'context': specify its structure (e.g., object with 'working_directory', 'environment_vars') or make it optional with clear guidance on when it's needed.
• Add error handling responses: document what errors the tool can return (timeout, permission denied, invalid input), whether they are retryable, and what recovery action the LLM should take.
• Add toolAnnotations to MCP tool registration: set readOnlyHint=true for agent_group-2, destructiveHint=true for agent_group-1 and agent_group-3 to align with declared risk levels.
• Verify tool definitions are explicitly registered in the main server code (src/cli/main.ts), not just in test fixtures. Test fixtures should verify behavior, not define tools.
• Add parameter validation rules in descriptions. Example: 'prompt (string, 1-2000 chars): The natural language task description for the agent to execute. E.g., "Read the file /tmp/config.json and extract the API key"'.