Two tools with clear verb-noun naming (enhance_prompt, critique_prompt) and reasonable descriptions (95-110 chars). Input schemas are present with Zod validation and proper types. However, output schemas are undocumented, responses return only text content with no structured fields documented for LLM planning. Parameter descriptions are adequate but lack format guidance. No error handling, recovery guidance, or output schema documentation. Descriptions are borderline short (below the 194-char baseline for A+ tools). Missing idempotency hints and composition guidance.

• Document output schema explicitly: 'Returns a text response containing [improved prompt | critique with questions]. The response is ready for the user to review or iterate on.' This helps LLMs understand what to do with the result.
• Expand tool descriptions to 150-200 chars. Add WHEN to use: 'Use enhance_prompt to rewrite rough drafts into polished, task-specific prompts. Use critique_prompt first if you need feedback before rewriting.' This disambiguates tool selection.
• Add runtime validation and error messages. Catch invalid target enum values and return: 'Invalid target: got "<value>". Must be one of: general, coding, image, writing, research.' This enables LLM self-correction.
• Document the 'context' parameter dependency: 'Optional. Provide domain-specific context (e.g., "for Claude 3.5 Sonnet") to tailor the rewrite. Improves relevance but not required.' This guides when to include it.
• Add idempotency note to both tools: 'Idempotent: calling with the same draft and target always produces the same guidance.' This signals safe retry behavior to agents.