Two well-named tools with complete input schemas and clear descriptions. Both tools follow verb_noun naming (format_currency, format_number) and have descriptions in the 100-200 char range, which is appropriate. Input schemas are properly typed with required fields declared. However, output schemas are not documented, LLMs cannot predict what fields to expect from responses. Error handling returns a generic error object with a hint, but lacks structured guidance for recovery. No parameter enums despite constrained inputs (style, currency codes). Missing descriptions for optional parameters (locale defaults to 'en-US' but this is buried in the description text, not formalized). Tools are read-only and offline, which is good for security, but no explicit documentation of this constraint in the schema.

• Document output schemas for both tools. Specify that success returns {amount, currency, locale, formatted} and errors return {error, hint}. Include field types and descriptions.
• Convert 'style' parameter to an enum: ['decimal', 'percent', 'compact']. This prevents hallucinated values and makes the tool self-documenting.
• Add a 'default' field to the 'locale' parameter schema: {type: 'string', description: '...', default: 'en-US'}. This formalizes the default and makes it machine-readable.
• Enhance error messages to guide recovery. Instead of generic 'Check the currency...', return specific errors like 'Invalid currency code: got "USDA", must be ISO 4217 (e.g., USD, EUR, JPY).'
• Consider adding optional parameters with constraints: for format_number, add min/max for 'decimals' (e.g., 0-20); for format_currency, document that 'amount' can be negative (for refunds).