Two tools with basic verb-noun naming (get_alerts, get_forecast). Both have short descriptions (40-50 chars) and documented input schemas with type and description fields. However, descriptions lack context on WHEN to use each tool, prerequisites, or what the output structure contains. No output schema documentation. Error handling returns plain strings without recovery guidance. Parameters are minimally described (state code, lat/lon) but lack constraints (e.g., state must be 2-letter code, lat/lon ranges). No pagination, no structured output, no tool annotations. Follows basic pattern:tool structure but misses pattern:tool-description depth and pattern:response-shaper rigor.

• Expand tool descriptions to 100-150 chars. Example: 'Get active weather alerts for a US state. Returns event type, severity, area, and instructions. Use this to check for warnings before get_forecast. State must be a 2-letter code (CA, NY, etc.).'
• Document output schema in tool docstrings. Example: 'Returns a formatted string with alert fields: Event, Area, Severity, Description, Instructions. If no alerts, returns "No active alerts for this state."'
• Add parameter constraints to descriptions. State: 'Two-letter US state code (CA, NY, TX, etc.). Invalid codes return an error.' Latitude/Longitude: 'Decimal degrees: latitude ±90, longitude ±180.'
• Implement structured error responses with recovery hints. Instead of 'Unable to fetch alerts', return 'Failed to fetch alerts for state code "XX": invalid state code. Valid codes are CA, NY, TX, ... Try a different state code.'
• Add tool annotations to FastMCP decorator: @mcp.tool(annotations={'readOnlyHint': True, 'idempotentHint': True}) to signal safety properties to LLMs.