Single tool 'stability.image' has a detailed description (194 chars) and comprehensive input schema with 6 parameters, all typed and described. However, the schema lacks critical constraints: no enum for 'output_format' (should restrict to png/jpeg/webp), no enum for 'size' (should list valid aspect ratios), and the 'n' parameter description contradicts itself ('must be 1' but accepts integer). Output schema is not documented, LLMs cannot predict response structure. Error handling and recovery guidance are absent from the tool definition. The tool name 'stability.image' is generic; 'generate_image' would be more action-oriented per verb_noun convention.

• Document the output schema: {image: string (base64), seed: integer, finish_reason: string}. This lets LLMs know what fields to extract and plan downstream operations.
• Convert 'output_format' to an enum: ['png', 'jpeg', 'webp']. Replace the description 'Optional output format (defaults to png)' with 'Output format. Defaults to png.'
• Convert 'size' to an enum listing all valid aspect ratios: ['21:9', '16:9', '3:2', '5:4', '1:1', '4:5', '2:3', '9:16', '9:21'] or WxH formats. Add constraint: 'If omitted, defaults to 1:1.'
• Remove parameter 'n' entirely if Stability always returns exactly one image per call. If variable counts are supported in future, add it back with proper enum [1] or remove the contradictory 'must be 1' clause.
• Rename tool from 'stability.image' to 'generate_image' to follow verb_noun convention. Update description to clarify: 'Generate images using Stability AI Stable Image (v2beta) API. Supports three model routes (core, sd3, ultra). Returns base64-encoded image with seed and finish_reason.'