Three tools with complete input schemas and descriptions, but descriptions are minimal (10-20 chars baseline) and lack LLM-optimized guidance. All tools are read-only, reducing composition complexity. Parameter descriptions are present but terse. No output schemas documented. Error handling and recovery guidance absent. Naming follows verb_noun pattern (console_*) but lacks clarity on distinctions between capture vs log vs screenshot.

• Expand tool descriptions to 100-150 characters. Example: 'Capture the current graphical console of a KubeVirt VMI as a PNG image. Use this to inspect the desktop UI when SSH is unavailable. Requires the VMI to be running and VNC-accessible.'
• Document output schemas for each tool. E.g., console_screenshot returns {image_data: base64_string, width: int, height: int, timestamp: ISO8601}; console_log returns {lines: string[], total_lines: int, truncated: boolean}.
• Add error handling examples in descriptions. E.g., 'If the VMI is not found, the tool returns error code 404 with available VMI names. If VNC is unavailable, returns 503 with guidance to check VMI status.'
• Enhance parameter descriptions with constraints and defaults. E.g., 'tail_lines: Number of lines from the end of the log (default: 500, max: 5000 to prevent memory exhaustion). Omit to retrieve the last 500 lines.'
• Clarify tool distinctions in descriptions. E.g., console_log: 'Retrieve historical serial console logs (up to 5000 lines).' console_capture: 'Stream live console output for a specified duration (up to 30s). Use this to monitor real-time activity.'