Bed of Roses provides 4 tools with complete input schemas and reasonable descriptions (avg 95 chars). All tools have type-annotated parameters with descriptions. However, descriptions lack LLM-optimized guidance on WHEN to use each tool vs alternatives, and output schemas are not documented. Error handling and recovery guidance are absent. Tool composition is sound (web_search → web_fetch chain is clear), but no pagination support for search results and no idempotency hints for the write operation (generate_image).

• Document output schemas for all 4 tools. Example for web_search: {results: [{title: string, url: string, snippet: string}], total_count: integer}. This enables LLMs to chain tools and extract data reliably.
• Add pagination to web_search: accept offset (default 0) and limit (default 5, max 10) parameters; return total_count and next_offset in response. Allows agents to iterate through large result sets without re-querying.
• Enhance tool descriptions with WHEN-to-use guidance. E.g., 'Use web_search to find relevant URLs and summaries. If you need full page content, call web_fetch on the URL.' This disambiguates tool selection.
• Add idempotency and confirmation hints to generate_image. Document that each call creates a new image (non-idempotent); consider adding a dry_run parameter or requiring explicit confirmation for production use.
• Document error scenarios and recovery paths. E.g., 'If web_fetch returns 404, the URL may be invalid or the page may have moved. Try searching for the content again with web_search.' This guides agent retry logic.