Server has 10 tools with basic schemas and descriptions, but significant quality gaps. All tools have descriptions (10-50 chars, mostly generic), and input schemas are present with type definitions. However, descriptions are too brief to guide LLM selection effectively (baseline 194 chars; these average ~30 chars). Parameter descriptions exist but lack context on constraints, formats, or when to use each tool. No output schemas documented. Error handling is minimal, HTTPException with generic messages. No tool annotations (readOnlyHint/destructiveHint). Naming is verb-forward but lacks clarity on distinctions (e.g., 'save_prompt_version' vs 'save_prompt_as_template' are confusingly similar). No pagination, no idempotency guidance, no recovery hints in errors.

• Expand all tool descriptions to 100-200 chars. Include WHAT (action), WHEN (use case), and any prerequisites. E.g., 'save_prompt_version: Creates a versioned snapshot of a prompt with a change note. Use this after editing a prompt to track changes over time. Requires an existing prompt_id and layer_prompt_id.'
• Document output schemas for all tools. Add a 'Returns' section to each tool description or metadata. E.g., 'Returns: {id: string, name: string, description: string, steps: array, created_at: ISO8601}'
• Add tool annotations: mark read-only tools with readOnlyHint=true (list_*, fetch_*, get_*) and destructive tools with destructiveHint=true (save_*, create_*, edit_*). This helps LLMs reason about side effects.
• Enhance error messages with recovery guidance. E.g., 'Template not found. Call list_prompt_templates() to see available templates, or create_layer_prompt() to create a new one.'
• Clarify naming: rename 'save_prompt_version' to 'create_prompt_version' to distinguish from 'save_prompt_as_template'. Use consistent verb_noun pattern across all tools.