Single tool with well-structured schema and reasonable description. Tool naming follows verb_noun convention (duckduckgo_web_search). Input schema is properly defined with type constraints, defaults, and enums. However, the description lacks strategic guidance on when to use this tool vs alternatives, and the output format is unstructured text rather than documented JSON. Rate limiting logic exists in code but is not exposed in tool documentation. No per-parameter descriptions beyond basic labels. Error handling returns text strings rather than structured error objects with recovery guidance.

• Document the return schema as a structured object: {results: [{title: string, description: string, url: string}], count: number, query: string}. Update performWebSearch() to return JSON instead of Markdown.
• Expand tool description to 150-200 chars with strategic context: 'Search using DuckDuckGo for general web queries, news articles, and diverse sources. Ideal for recent events and privacy-conscious results. Limited to 20 results per request. See also: search_google, search_academic for specialized searches.'
• Add per-parameter descriptions: safeSearch → 'Content filtering level: strict (blocks adult/explicit content), moderate (blocks explicit only), off (no filtering). Default: moderate.'
• Implement structured error responses: {error: string, code: 'RATE_LIMIT_EXCEEDED'|'INVALID_QUERY'|'NO_RESULTS', retryable: boolean, guidance: string}. Example: {error: 'Rate limit exceeded', code: 'RATE_LIMIT_EXCEEDED', retryable: true, guidance: 'Wait 60 seconds before retrying; monthly quota resets daily at UTC 00:00.'}
• Document rate limits in the tool description: 'Rate limited to 1 request/second and 15,000/month.'
• Add an optional 'region' parameter (enum: us, uk, de, fr, etc.) if DuckDuckGo API supports it, with description: 'Region for localized results. Default: us.'
• Include sample query-result mapping in docs to show agents what output structure looks like, not in the description itself (to avoid LLM value hallucination).