marine-mcp demonstrates solid naming conventions (all tools start with action verbs: search_, get_, lookup_, validate_, obis_) and comprehensive descriptions (avg 150+ chars). However, parameter descriptions are inconsistent, some parameters lack descriptions entirely (e.g., region_id in get_region_briefing, region_id in get_active_missions, difficulty in get_active_missions). Input schemas use Zod with proper types, but output schemas are not documented in the code. Error handling is minimal, tools return generic envelope structures without actionable recovery guidance. The review-gating pattern is well-articulated in tool descriptions, but parameter constraints (e.g., size limits for obis_gbif_passthrough) lack explicit documentation in descriptions.

• Add descriptions to all parameters: region_id in get_region_briefing ('e.g., "atlantic-north", "pacific-west"'), region in get_active_missions, difficulty in get_active_missions. Follow pattern: 'The [resource] [constraint], e.g., [example]'.
• Document output schemas explicitly. Add a comment block above each tool handler showing the response structure, e.g., '{ id, title, species_group, status, review, servable }' for search_species.
• Enhance error responses with recovery hints. When species_id not found, return: 'Species not found. Try search_species() with a partial name or region.' When servable:false, suggest: 'This artifact is under review. Check back later or contact the commons maintainers.'
• Move numeric constraints into parameter descriptions. For obis_gbif_passthrough size param, change description to: 'Number of records to return (1 - 20, default 10).'
• Consider adding a batch_get_species_details tool accepting an array of species_ids, returning results with per-item success/failure. This reduces token waste for agents fetching multiple species.