Two tools with explicit schemas and descriptions. bash tool has clear, actionable description (71 chars) and well-structured input schema with two parameters (command: required string, run_in_background: optional boolean). ask_user_question has comprehensive description (287 chars) and detailed nested schema for questions array with proper constraints (1-4 items, 2-4 options per question). However, both tools lack documented output schemas, critical for LLM planning. Parameter descriptions are present but minimal (e.g., 'The shell command to run' could specify format/constraints). No error handling guidance visible. Risk annotations (DESTRUCTIVE, READ_ONLY) are present but not formalized as tool annotations in the MCP sense.

• Document output schemas for both tools. bash should return {exit_code: number, stdout: string, stderr: string, bg_id?: string}. ask_user_question should return {answers: array of {question: string, selected_option: string}}.
• Add error handling guidance to bash: 'If exit_code != 0, stderr contains the error. Timeout errors are retryable; permission errors require user intervention.'
• Implement input validation for bash: whitelist safe commands or reject dangerous patterns (rm -rf, dd, mkfs). Document constraints in parameter description.
• Add a dry_run parameter to bash (optional boolean) to preview command output without side effects, reducing accidental destruction.
• Clarify ask_user_question behavior: document that 'Other' option is auto-appended, explain when to use multiSelect=true, and specify response format for custom text.

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

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