Server has 4 tools with explicit schemas and descriptions, but quality is significantly below production baseline. All tool descriptions are in Chinese and lack English translations, limiting LLM accessibility. Descriptions are present but minimal (10-30 chars), below the 50-200 char LLM-optimized baseline. Parameters lack descriptions entirely, 'file_path' has a description, but other tools have empty properties objects with no parameter guidance. No output schemas documented. Error handling is minimal (generic error messages). No tool annotations (readOnlyHint, destructiveHint, idempotentHint). ipfs_upload is a WRITE operation but lacks confirmation/dry-run pattern. No pagination support despite potential for large result sets (ipfs_pins, cluster_peers). Tool composition is reasonable (single responsibility), but parameter quality is poor.

• Translate all tool descriptions and parameter descriptions to English. Use clear, action-oriented language: 'Upload a local file to IPFS via starstack.store and return the content hash (CID) and a public gateway URL.' Aim for 50-200 characters.
• Add output schema documentation for each tool. Example for ipfs_upload: 'Returns {cid: string, gateway_url: string, ...}'. For ipfs_pins: 'Returns {pins: [{cid: string, name: string, size: number}], total: number}'.
• Add destructiveHint annotation to ipfs_upload in the tool definition: 'destructiveHint': true. This signals to LLMs that the operation has side effects.
• Add pagination parameters (limit, offset/cursor) to ipfs_pins and cluster_peers. Document max result limits in descriptions: 'Returns up to 50 pinned items. Use offset/limit for pagination.'
• Enhance error messages with recovery guidance. Instead of generic 'starstack.store POST /api/ipfs-upload failed: ...', return 'File upload failed: check file exists and is readable. Verify STARSTACK_BASE_URL is set correctly.'