Four tools with complete input schemas and descriptions, but quality varies significantly. All tools have descriptions (10-200 chars range) and typed parameters with descriptions. However, descriptions lack LLM-optimized guidance on WHEN to use each tool vs alternatives, and error handling is minimal. Output schemas are not documented. Parameter descriptions are functional but sparse (avg ~60 chars vs baseline 72). No tool annotations (readOnlyHint/destructiveHint/idempotentHint) despite all being WRITE operations. Naming is clear and verb-forward (save_*, log_*, check_*), but composition could be tighter, save_listing and log_listing_check are near-duplicates with different outcomes.

• Add tool annotations: mark all four tools with destructiveHint=true and idempotentHint=false (or true if retries are safe). This tells LLMs which tools can be safely retried.
• Document output schemas for each tool. E.g., save_listing returns {listing_id: integer, created_at: string}. This enables downstream tool chaining.
• Clarify the distinction between save_listing and log_listing_check in their descriptions. E.g., 'save_listing: persist a matching listing for future price checks. log_listing_check: record rejection reasons without saving.'
• Add min/max constraints to numeric parameters in schema. E.g., match_score: {type: integer, minimum: 0, maximum: 100}. Repeat constraints in descriptions.
• Add error handling guidance to each tool description. E.g., 'If URL is not on the target site, returns error with allowed domain. If price format is invalid, returns error with expected format (decimal string).' This guides LLM recovery.