Two tools with basic schemas and descriptions. Both tools have input schemas (one with optional parameter, one empty), descriptions present but minimal (10-20 chars), and tool annotations (readOnlyHint). However, parameter descriptions are sparse, output schemas are not documented, and error handling guidance is absent. The 'start' tool accepts an optional 'name' parameter with a description, but 'get-fortune-cookie' has no parameters documented. No pagination, no error recovery patterns, no output field documentation. Meets baseline for a demo but lacks production-grade rigor.

• Expand tool descriptions to 50-200 characters. Example: 'start' → 'Initiate Skybridge onboarding flow. Accepts optional user name to personalize the experience. Returns structured onboarding state for the view to render.'
• Document output schemas explicitly. For 'start', specify: returns {structuredContent: {name?: string}, content: [{type: 'text', text: string}], isError: boolean}. For 'get-fortune-cookie', specify: returns {structuredContent: {prediction: string}, ...}.
• Add error handling examples in descriptions. E.g., 'If onboarding fails, the LLM should retry with a simpler name or ask the user for clarification.'
• If 'get-fortune-cookie' has optional parameters (e.g., language, category), declare them in the schema with descriptions and enums.
• Add per-tool audit logging and permission checks. Document what permissions each tool requires (e.g., 'read:user_profile', 'write:onboarding_state').