Two tools with basic descriptions and partial schemas. list-prompts and get-prompt both have action verbs and descriptions (10-20 chars each), but descriptions are minimal and lack context on when/why to use them. Input schemas present with type info but parameter descriptions are generic. No output schema documented. Error handling exists but lacks recovery guidance. Tool composition is reasonable (list then get pattern), but parameter naming could be clearer (e.g., 'filter' vs 'searchFilter').

• Expand tool descriptions to 50-150 chars. Example: 'list-prompts: Retrieve all prompts stored in MLflow Prompt Registry. Returns paginated list with prompt names and template variables. Call this first to discover available prompts before using get-prompt.'
• Document output schemas in tool descriptions or as structured comments. Example: 'Returns {prompts: [{name: string, arguments: [{name: string, required: boolean}]}, nextCursor?: string]}'
• Add format details to parameter descriptions. For 'filter': 'Search filter using ILIKE syntax (case-insensitive substring match). Example: "my_prompt" matches "my_prompt_v1", "my_prompt_v2".' For 'arguments': 'JSON object mapping template variable names to values. Variable names extracted from {{variable}} syntax in prompt template.'
• Implement error classification and recovery hints. Catch specific MLflow API errors (404 → 'Prompt not found. Call list-prompts to see available prompts.'; 500 → 'MLflow service unavailable. Retry in 30 seconds.'; network timeout → 'Connection timeout. Verify MLFLOW_TRACKING_URI is reachable.')
• Add parameter validation with actionable errors. For 'arguments', validate keys match extracted template variables and return: 'Invalid argument keys: got ["foo", "bar"], expected ["user_name", "context"] from prompt template.'