Two tools with clear, concise descriptions and basic input schemas. Tool names follow verb_noun pattern (rastrear_*). Input schemas present with type definitions and descriptions. However, output schemas are undocumented, LLMs cannot predict response structure. Parameter descriptions lack format constraints (e.g., regex pattern for codigo not stated in schema description). Error handling returns structured objects but lacks recovery guidance. No tool annotations (readOnlyHint, idempotentHint). Validation logic exists in code but not exposed in schema constraints.

• Document output schemas for both tools. Specify fields (e.g., {"erro": boolean, "mensagem": string, "objeto": {...}, "eventos": [...]}) so LLMs know what to extract.
• Add format constraint to 'codigo' parameter description: 'Tracking code in format: 2 uppercase letters + 9 digits + 2 uppercase letters (e.g., AA000000000BR). Invalid format returns error.'
• Add maxItems: 20 constraint to 'codigos' array schema and restate in description: 'Array of tracking codes (max 20 per call).'
• Enhance error messages with recovery hints. E.g., invalid code: 'Invalid code format. Expected: 2 letters + 9 digits + 2 letters (e.g., AA000000000BR). Check the tracking label on your package.'
• Add tool annotations to schema: set readOnlyHint: true for both tools to signal they are safe to call without side effects.