Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and error handling. get_task has optional parameters with minimal guidance on defaults and dependencies. check_pricing lacks clarity on what 'model' and 'action' mean in context. Neither tool documents output structure or error recovery paths. Descriptions are present but generic (10-20 chars above minimum). No tool annotations (readOnlyHint, idempotentHint). Parameter descriptions exist but are terse and lack format/constraint guidance. No pagination, no batch operations, no composition hints.

• Add explicit enum constraints for 'action' parameters. Instead of free-form strings, use z.enum([...endpoints]) to make valid options discoverable and prevent hallucinated values.
• Document output schemas in tool descriptions. E.g. 'Returns {task_id, status, task: {id, status, result_url, created_at}}' so LLMs know what fields to extract.
• Enhance error messages with recovery hints. Change 'Unsupported RunAPI service/action/model combination' to 'Unsupported model. Available models: gpt-4o-image. Verify the model parameter matches the line.'
• Add readOnlyHint and idempotentHint annotations to both tools via server.tool() options to signal retry safety and lack of side effects.
• Expand parameter descriptions with format guidance. E.g. 'action: Asynchronous endpoint name (e.g., gpt_4o_image_generate). Defaults to the line's only endpoint if only one exists.'