Codesign-mcp has 8 tools with clear action-verb naming (get_, list_, download_, debug_) and reasonable descriptions (avg ~100 chars). However, critical gaps exist: parameter descriptions are sparse or missing entirely (e.g., 'targetPlatform' in get_artboard_spec lacks guidance on valid values; 'customScale' has no range or units). Output schemas are not documented, LLMs cannot predict what fields to expect from responses. Error handling is absent from descriptions. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. The server is functional but falls short of production-grade tool design.

• Add output schema documentation for each tool. Example: list_artboards should document {artboards: [{name: string, id: string, width: number, height: number}], total: number}. Include all IDs needed for downstream calls.
• Convert 'targetPlatform' parameter to an enum: ['web', 'ios', 'android', 'wechat-miniprogram']. Add bounds to numeric params: customScale (0.1 - 5.0), customWidth (100 - 4000), remBasePx (8 - 32).
• Add tool annotations to server.ts ToolConfig. Mark codesign_login/logout with destructiveHint=true, read-only tools with readOnlyHint=true, idempotent tools with idempotentHint=true.
• Enhance descriptions with error recovery guidance. E.g., 'If login fails, check browser connectivity. If sharing URL is invalid, verify the URL format (https://codesign.qq.com/s/...).'
• Add pagination to list_artboards: limit (1 - 100, default 20) and offset (default 0) parameters. Return {artboards: [...], total: number, hasMore: boolean} to enable agent-driven pagination.

Inferred effective spec: 2025-06-18+.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry