Two tools with basic verb-noun naming (get_alerts, get_forecast). Both have short descriptions (34-45 chars, below the 50-200 char baseline for LLM optimization). Input schemas are present with typed parameters and descriptions, but output schemas are completely undocumented, LLMs cannot plan downstream calls or extract structured data. No error handling guidance, no parameter constraints (e.g., state code validation), and no indication of pagination or result limits. Tools return free-text strings rather than structured objects, forcing LLMs to parse unstructured output.

• Expand tool descriptions to 80-150 chars. Example: 'Get active weather alerts for a US state. Returns event type, severity, area, and instructions. Call this to check for warnings before get_forecast.'
• Document output schemas explicitly. Define return type as structured object: {alerts: [{event, area, severity, description, instructions}]} or {forecast: [{period, temperature, wind, forecast}]}.
• Add enum constraint to 'state' parameter: state must be a 2-letter US state code (CA, NY, TX, etc.). Validate in code and return 'Invalid state code: got "XX", must be 2-letter code like CA, NY.'
• Add error recovery guidance. Replace 'Unable to fetch alerts' with 'Failed to fetch alerts for state {state}. Verify state code is valid (e.g., CA, NY). If error persists, the NWS API may be unavailable.'
• Add limit and pagination support. Accept optional 'limit' parameter (default 20, max 100) and return {alerts: [...], total_count: N, has_more: bool}. Document in description: 'Returns up to 20 alerts; use limit parameter to adjust.'