Two tools with basic schemas and descriptions, but significant gaps in parameter documentation, output schema clarity, and error handling guidance. search_plots has 11 boolean parameters with minimal descriptions (e.g., 'Filter for plots near coastline', no context on what 'near' means or distance thresholds). list_prefab_homes lacks output schema documentation. Both tools return raw JSON without structured field definitions. Error handling returns generic messages without recovery guidance. No tool annotations (readOnlyHint, idempotentHint). Descriptions are adequate (50-100 chars) but lack LLM-optimized context on when to use each tool vs. the other.

• Add explicit output schema documentation as JSDoc comments above each tool handler. Example: '// Returns: { id, title, price_eur, size_m2, listing_url, ... }' with field types and presence guarantees.
• Expand parameter descriptions with constraints and examples. For 'near_coastline', specify: 'Boolean. If true, returns only plots within 5km of coastline (based on enrichment_data). Mutually compatible with other near_* filters.'
• Implement pagination: add 'offset' (default 0) and 'limit' (default 20, max 100) parameters to both tools. Return { results: [...], total_count, has_more } to enable agent-driven pagination.
• Enhance error messages with recovery hints. Example: 'No plots found matching filters. Try: (1) increasing max_price_eur, (2) removing near_* filters, or (3) calling search_plots with no filters to see available data.'
• Add tool annotations (readOnlyHint: true, idempotentHint: true) to both tools since they are read-only queries with no side effects. This signals to agents that retries are safe.