Knovaryn has 3 tools with basic schemas and descriptions, but significant gaps limit production readiness. The health and knovaryn_doctor tools are READ_ONLY diagnostics with empty input schemas ({}), which is acceptable for parameterless tools. knovaryn_create_project has a proper schema with 5 parameters, all typed and described. However, descriptions are minimal (10-50 chars), falling below the 50-200 char baseline for LLM-optimized tools. No output schemas are documented, no error handling guidance is provided, and no tool annotations (readOnlyHint/destructiveHint) are visible. The create_project tool lacks enum constraints on tags and no validation rules are stated for slug format or owner_principal. Parameter descriptions lack actionable constraints (e.g., slug length, character restrictions). No recovery guidance for common failures (e.g., 'slug already exists').

• Expand tool descriptions to 50-200 characters. For health: 'Check server health and identity. Returns version, status, and capabilities.' For knovaryn_doctor: 'Run environment diagnostics (Python version, dependencies, optional extras). No secrets exposed.'
• Document output schemas for all tools. E.g., health returns {version: string, status: string, capabilities: [string]}; knovaryn_create_project returns {project_id: string, default_profile: string, actions: [string]}.
• Add constraints to knovaryn_create_project parameters: slug as regex pattern (e.g., '^[a-z0-9_-]{3,50}$'), tags as array of strings with max 10 items, owner_principal format documented.
• Add error handling guidance: 'If slug already exists, try appending a timestamp or UUID. If owner_principal is invalid, call list_principals() first to find valid IDs.'
• Add tool annotations: mark health and knovaryn_doctor with readOnlyHint=true; mark knovaryn_create_project with destructiveHint=true (creates persistent resource).