Server provides 10 tools with consistent naming (verb_noun pattern) and reasonable descriptions (avg ~100 chars). All tools have input schemas with type definitions. However, parameter descriptions are sparse or generic, output schemas are undocumented, and error handling lacks recovery guidance. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. Descriptions meet minimum length but lack LLM-optimization details (e.g., when to call, dependencies, expected output structure).

• Document output schemas for all tools. Example: 'Returns {components: [{name: string, category: string, description: string}], total: number}'. This enables agents to chain tools and extract the right fields.
• Add tool annotations to capabilities. Mark read-only tools with readOnlyHint=true; mark apply_theme with destructiveHint=true. This signals to agents which tools are safe to retry and which require caution.
• Expand parameter descriptions with format/constraint details. E.g., 'componentName: Name of the shadcn/ui component in kebab-case (e.g., "accordion", "button"). Case-sensitive. Call list_components() first if unsure.'
• Add error recovery guidance. E.g., 'If component not found, call list_components() to see available options. Common typos: "button-group" → "button", "input-field" → "input".'
• Make apply_theme safer by defaulting dryRun=true and requiring explicit dryRun=false to write. Or add a separate preview_theme tool that returns the changes without applying them.

Inferred effective spec: <=2025-11-25.

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