mcp-hfspace has 3 tools with significant definition gaps. Two static tools (available-files, search-spaces) have descriptions but lack proper parameter documentation and output schemas. The third tool (dynamic_endpoint_tools) is inferred rather than explicitly defined in the source, making it impossible to assess its actual schema. Parameter descriptions are missing or trivial across all tools. No output schemas are documented. Error handling exists but lacks recovery guidance.

• Rename 'available-files' to 'list_available_files' to follow verb_noun naming convention.
• Add explicit output schema documentation to all tools. For available-files, define the markdown table structure with fields like: uri (string), name (string), size (number), lastModified (ISO 8601 string), mimeType (string).
• Expand search-spaces query parameter description to specify: valid length (3-7 words recommended), examples of tasks (e.g., 'image classification', 'text-to-speech'), and what the markdown table columns represent.
• Document output schemas inline in tool definitions. E.g., for search-spaces: 'Returns a markdown table with columns: SpaceName | Description | Author | URL | Tags.'
• Add error classification to search-spaces error handling: 'Search failed (retryable). Try a shorter, more specific query or contact Hugging Face support.'
• For dynamic_endpoint_tools, make endpoint definitions explicit in the source: register each Gradio endpoint with its full name, description, parameters, and output schema in a structured way (not inferred from endpoint_wrapper).
• Add pagination guidance for tools that may return many results: 'If more than 20 results are available, implement limit and offset parameters to prevent context overflow.'
• Include example invocations in tool descriptions to clarify expected usage patterns.
• Document which tools are read-only and which have side effects. Currently only inferred from 'Risk' tags (READ_ONLY, WRITE) in the server description.