Two tools with basic verb-noun naming (get_alerts, get_forecast). Both have descriptions and input schemas with typed parameters. However, descriptions are minimal (under 100 chars) and lack context on when to use each tool or what to do with results. No output schema documentation. Parameters lack constraints (e.g., state code validation, lat/lon bounds). Error handling returns plain strings with no recovery guidance. No tool annotations. Schemas are present but sparse, missing examples, constraints, and output structure documentation.

• Add output schema documentation. Define the structure returned by each tool (e.g., get_alerts returns array of {event, area, severity, description, instructions}; get_forecast returns array of {name, temperature, wind, forecast}). This enables LLM chaining and structured parsing.
• Add parameter constraints. For 'state', use enum of valid US state codes or a regex pattern. For latitude/longitude, document bounds (lat: -90 to 90, lon: -180 to 180) and precision (e.g., 4 decimal places = ~11m accuracy).
• Expand descriptions to 150 - 250 chars. Include: what the tool does, when to use it (e.g., 'Call this to check for active weather warnings before travel'), and what it returns. Example: 'Fetch active weather alerts for a US state. Returns event type, affected area, severity, and instructions. Use this to warn users of hazards like tornado watches or flood advisories.'
• Improve error messages with recovery guidance. Instead of 'Unable to fetch alerts', return 'Failed to fetch alerts for CA (API timeout). Try again in 30 seconds or check https://weather.gov for manual updates.' This guides the LLM's next action.
• Add tool annotations. Mark both tools with readOnlyHint=true and idempotentHint=true so agents know they are safe to retry and do not modify state.