Three tools with basic schemas and descriptions, but significant gaps in parameter documentation, output schema clarity, and error handling guidance. Tool names follow verb_noun convention (generate_text, analyze_text, chat), which is good. However, descriptions lack depth on WHEN to use each tool vs. alternatives, and parameter descriptions are minimal. No output schemas documented. Error responses are generic HTTP status codes with no recovery guidance. Missing idempotency hints, rate-limit guidance, and dependency documentation.

• Document output schemas for all three tools. E.g., generate_text returns {text: string, model: string, usage: {prompt_tokens: int, completion_tokens: int, total_tokens: int}}. This lets LLMs plan downstream calls.
• Enhance parameter descriptions with constraints and examples. E.g., 'temperature (0.0 - 1.0, default 0.7): Controls randomness. Use 0.0 for deterministic summaries, 0.7 for creative writing, 1.0 for brainstorming.'
• Add WHEN-to-use guidance to tool descriptions. E.g., 'generate_text: One-shot text generation. Use for single prompts without conversation history. For multi-turn chats, use chat() instead.'
• Implement structured error responses with recovery hints. E.g., {error: 'Missing required parameter: prompt', hint: 'Provide a string 1-10000 characters. Example: "Write a haiku about AI."', retryable: false}
• Add rate-limit and timeout warnings to descriptions. E.g., 'Gemini API has rate limits. If you hit a 429 error, wait 60 seconds before retrying. Calls timeout after 30 seconds.'