Single tool with adequate naming but significant gaps in description quality and schema documentation. Tool name 'get_nice_guidance' follows verb_noun convention (+). Description is 194 chars, within baseline range, but lacks critical context: no mention of prerequisites (SERPER_API_KEY required), no guidance on when to use vs alternatives, no output structure documented. Parameter 'query' has a description but lacks constraints (format, length, examples of valid queries). No output schema documented, LLM cannot predict response structure. No error handling guidance (what happens on API failure, timeout, no results). Tool performs external API calls with 30s timeout but no timeout error recovery documented.

• Document output schema: specify that tool returns a single string containing concatenated guidance text, or refactor to return structured object with fields like {guidance: string, sources: [{url, title, snippet}], total_results: int}.
• Add error handling to description: 'Returns empty string if no results found or API timeout occurs. Requires SERPER_API_KEY environment variable.'
• Constrain query parameter: add description 'Medical topic query (2-100 chars, e.g. "high blood pressure", "diabetes management"). Searches NICE guidance only.'
• Implement result limiting: cap concatenated text to 4000 chars and document in description. Add pagination support (offset, limit params) for future expansion.
• Add prerequisite documentation: 'Requires SERPER_API_KEY from https://serper.dev. Set via environment variable before server startup.'