Single tool 'call_dk' has critical definition gaps. Description is only 4 characters ("呼叫dk"), far below the 10-character minimum and provides zero context for LLM selection. Input schema is completely empty ({}), no parameters defined, no types, no descriptions. Tool appears to be a wrapper around a GUI subprocess that returns mixed text and images, but this is entirely undocumented. No error handling guidance, no output schema documentation, no parameter constraints. The tool name 'call_dk' is opaque, does not start with a clear action verb and provides no indication of what 'dk' means or what the tool does. Code shows it launches a PySide6 GUI and processes results, but none of this is reflected in the tool definition.

• Expand tool description to 50-200 characters explaining what 'dk' is, what the tool does, when to call it, and what it returns. Example: 'Launch the DK interactive assistant UI to generate code, debug, or analyze files. Returns text response and optional screenshots. Use when user needs interactive code generation or visual debugging.'
• Define explicit input parameters in schema: 'project_directory' (string, required, description: 'Path to project root for context'), 'summary' (string, optional, description: 'User prompt or task description for DK assistant'). Add type, format, and constraints.
• Rename tool to a clearer verb_noun pattern: 'launch_dk_assistant', 'invoke_dk_ui', or 'run_dk_interactive', something that signals the action and what 'dk' is.
• Document output schema: 'Returns object with fields: text (string, combined user input and command logs), images (array of Image objects with format and data). May return empty text if only images generated.'
• Add error handling guidance: 'If UI launch fails, check project_directory exists and is readable. If image processing fails, verify PIL/Pillow is installed. Retry with simpler prompt if timeout occurs.'

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry