Zhuling exposes 8 tools with minimal documentation. Tool names lack action verbs (execute, executeWithMetadata, registerLocalTool, registerMcpTools, registerSkillsTools, getAllToolCallbacks, getToolCallback, getRegisteredToolNames). Descriptions are present but generic (10-100 chars, well below the 194-char baseline). Input schemas visible for all tools but lack depth: parameters have types and descriptions, but no enums, constraints, or validation guidance. Output schemas are not documented. Error handling is absent, no recovery guidance, categorization, or actionable messages. The tools appear to be internal framework APIs rather than user-facing agent tools, which explains the sparse documentation. No security considerations visible (credentials, permissions, audit trails).

• Rename tools to follow verb_noun convention: 'execute' → 'run_tool', 'registerLocalTool' → 'register_local_tool', 'getAllToolCallbacks' → 'list_tool_callbacks', 'getToolCallback' → 'get_tool_callback', 'getRegisteredToolNames' → 'list_registered_tools'.
• Expand descriptions to 150 - 250 characters. For 'execute': 'Executes a registered tool with the specified input parameters. Returns the tool's output or an error. Use this to invoke any registered tool by name. Requires the tool to be registered first via register_*_tool.'
• Document output schemas for all tools. E.g., 'execute' returns {success: boolean, output: object, duration_ms: number, error?: string}. 'list_tool_callbacks' returns {tools: [{name: string, description: string, input_schema: object}], total: number}.
• Add error handling guidance: 'If tool not found, call list_registered_tools() first. If execution fails, check tool input format matches the tool's schema. Retryable errors: timeout, rate_limit. Non-retryable: invalid_tool_name, malformed_input.'
• Constrain 'toolInput' parameter: 'Valid JSON string (max 10KB). Must match the target tool's input schema. Example: {"key": "value"}. Validate before calling execute.'.