Three tools with well-structured schemas and detailed descriptions. web_search and fetch_url have comprehensive parameter documentation with enums and freshness hints. save_finding is write-capable with clear intent. All tools have input schemas with type definitions and descriptions. However, output schemas are not documented in the visible code, and error handling guidance is minimal. Descriptions are strong (150-250 chars, well above the 10-char minimum), but lack explicit recovery hints for failure cases.

• Document output schemas for all three tools. For web_search, specify: {results: [{title: string, url: string, excerpt: string, cache_age_seconds?: number}], cache_header?: string}. For fetch_url: {content: string, truncated: boolean, url: string}. For save_finding: {success: boolean, query: string, cached_at: string}.
• Add error recovery hints to each tool description. Example for web_search: 'If results are empty or irrelevant, try a more specific query or set force_fresh=true. If you need a list, set full_results=true.'
• Clarify parameter dependencies in web_search: 'If force_fresh=true, freshness is ignored. Use force_fresh for live data (prices, scores) and freshness for cache tuning.'
• Document fetch_url budget and timeout: 'Returns up to 8000 characters of extracted text. Pages longer than this are truncated with [... truncated ...].' Add timeout guidance: 'If fetch times out after 30s, the URL may be slow or unreachable.'
• Add permission/scope declarations to tool descriptions. Example: 'web_search requires read-only access to public web. save_finding requires write access to local cache.' This enables least-privilege agent configuration.