This server has significant definition quality gaps. Tool names lack action verbs (e.g., 'stats', 'models', 'nodes' instead of 'get_stats', 'list_models', 'describe_nodes'). Most parameter descriptions are minimal (10-30 chars), well below the 72-char baseline. No output schemas are documented. Error handling is absent, tools fail silently or with raw HTTP errors. The 'submit' tool combines workflow queuing, polling, and downloading into one operation, violating single-responsibility. Input schemas are present but sparse; many lack type constraints or enums. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels.

• Rename tools to verb_noun pattern: get_stats, list_models, describe_nodes, queue_workflow, poll_workflow, download_outputs, upload_image, interrupt_generation, free_memory.
• Expand parameter descriptions to 50-150 chars. E.g., 'type' → 'Model category (checkpoints, loras, vae, diffusion_models, text_encoders, clip, controlnet, upscale_models). Required.' 'timeout' → 'Max seconds to wait for workflow completion (1-3600, default 1800).'
• Document output schemas for all tools. E.g., models returns {type: string, items: [{name: string, size_bytes: integer}]}. submit returns {prompt_id: string, outputs: {[node_id]: {images: [{filename, subfolder, type}]}}}.
• Split submit into three tools: queue_workflow (returns prompt_id), poll_workflow (returns status + outputs), download_outputs (returns file paths). Enables agent to compose and retry independently.
• Add error recovery hints. E.g., models: 'If type is invalid, call list_model_types() first.' submit: 'If workflow fails, check /history/{prompt_id} manually or call interrupt_generation() to free resources.'