Nova exposes 5 tools with significant definition gaps. Tool names lack clear action verbs (ask_user, delegate_to_agent are vague; code_run is generic). Descriptions are present but minimal (10-50 chars), below the 50-200 char LLM-optimized baseline. Input schemas are partially visible in frontend code but lack complete type definitions and parameter descriptions in the backend. The code_run tool accepts arbitrary Python with no validation or safety constraints. edit and write tools lack content parameter schemas. No output schemas documented. Error handling is absent, no recovery guidance or categorization.

• Rename code_run to execute_python_code and add mandatory description: 'Execute Python code in a sandboxed environment. Returns stdout, stderr, and exit code. Supports only safe stdlib modules; blocks file I/O and network access.'
• Add complete input schemas for edit and write tools with: filePath (string, required, path validation), content (string, required, max 1MB), encoding (enum: utf-8|ascii, default utf-8), createIfMissing (boolean, default false).
• Document output schemas for all tools. E.g., code_run returns {stdout: string, stderr: string, exit_code: int, execution_time_ms: int}.
• Expand tool descriptions to 80-150 chars following pattern: '[VERB] [OBJECT]. [WHEN/WHY]. [RETURNS].' E.g., 'Execute Python code in a sandboxed environment. Use for calculations, data processing, or testing. Returns stdout, stderr, and exit code.'
• Implement input validation and error recovery guidance. E.g., code_run should catch SyntaxError and return 'Invalid Python syntax at line X: [details]. Check indentation and quotes.' instead of raw tracebacks.