WILDSCAN has 15 tools with mostly complete schemas and descriptions, but significant gaps in parameter documentation and error handling guidance. Tool names follow verb_noun convention well (list_detections, analyze_text, score_listing). Descriptions average ~120 chars, meeting the 10-1024 baseline. However, 8 tools lack parameter descriptions for complex object inputs (e.g., analyze_images accepts 'listing' object with no field documentation; classify_species accepts 'db' parameter with no guidance). Output schemas are not documented in the provided code. Error handling is absent, no recovery guidance, retryability classification, or actionable error messages. Security concerns: review_detection accepts 'reviewer' parameter with default 'investigator', but no permission gating visible. The WebSocket tool (detection_feed) has no input schema or error handling documented.

• Document nested object schemas for analyze_text (listing fields: title, description, images_local, images), analyze_images (listing fields), classify_species (text_analysis, image_analysis, listing, db fields), and score_listing (signals subfields: code_word_confidence, cites_appendix, iucn_status, etc.). Use JSON Schema $ref or inline definitions.
• Add output schema documentation to all tools. Specify return type (object, array, string), key fields, and field types. E.g., list_detections returns {detections: [{id, tier, species, risk_score, ...}], total, limit, offset}.
• Implement error handling with recovery guidance. Return structured errors: {error_code, message, retryable, next_steps}. E.g., 'species_not_found: Try classify_species with different image or text analysis.'
• Add permission checks to review_detection and download_images. Verify caller has 'write:detections' or 'write:images' scope before executing. Log who reviewed/downloaded what.
• Document detection_feed WebSocket event schema. Specify event types (detection_found, scan_complete, error), required fields per event, and reconnection behavior.