22 tools with minimal documentation. All tools have basic descriptions (10-50 chars), but lack depth on WHEN to use them, prerequisites, or return value structure. No visible input schemas in source code, parameter types and descriptions are inferred from the tool list provided, not from actual schema definitions. No output schemas documented. Error handling guidance absent. Tool names are action-oriented (execute, kernel_reset, upload) but descriptions are terse and do not explain dependencies, side effects, or expected outcomes. No parameter validation rules, enums, or constraints visible. This server reads like a thin wrapper over Colab APIs without LLM-optimization.

• Add explicit JSON Schema definitions for all 22 tools. Include input and output schemas with type declarations, required fields, and descriptions for every parameter and response field.
• Expand tool descriptions to 50-200 characters. For each tool, explain: (1) what it does, (2) when to use it instead of similar tools, (3) what it returns, (4) any prerequisites or side effects. E.g., 'Execute Python code on the warm Colab kernel. Returns stdout, stderr, and execution result. Timeout in seconds (1-3600). Use execute for synchronous code; execute_detached for background jobs.'
• Document parameter constraints in descriptions. For execute: 'code (string, max 100KB), timeout (integer, 1-3600 seconds, default 30)'. For expose: 'port (integer, 1-65535, must be available locally)'.
• Add error handling guidance to destructive tools. E.g., kernel_reset: 'Resets the kernel state. This is irreversible. Returns error if kernel is busy; wait 5 seconds and retry. On success, all variables are cleared.'
• Include output schema examples or field descriptions. E.g., kernel_info should document: 'Returns {kernel_id, status, python_version, memory_usage_mb, uptime_seconds}'. This enables LLMs to chain tools correctly.