Server defines 5 tools with clear verb-noun naming (list_*, get_*) and reasonable descriptions (avg 180 chars). All tools have input schemas with type declarations. However, critical gaps: (1) output schemas are completely undocumented, LLMs cannot plan downstream calls or extract fields; (2) parameter descriptions lack format/constraint details (e.g., 'goal' accepts '1-17' but description doesn't specify validation); (3) no error handling guidance, API failures return raw HTTP errors with no recovery hints; (4) no pagination support despite list_* tools potentially returning large datasets; (5) tool descriptions reference external series codes ('SI_POV_DAY1') without explaining how to discover them. Naming is solid (verb_noun pattern, action-clear), but schema and error handling are production gaps.

• Document output schemas for all tools. Example: 'Returns { data: [{ goal: string, title: string, description: string, targets?: [...] }] }' for list_goals.
• Add error handling with recovery hints. Wrap sdgGet() to catch HTTP errors and return: 'Series code not found. Call list_series() to see available codes.' instead of raw 500 errors.
• Add enum constraint to 'goal' parameter: enum: ['1', '2', ..., '17'] or pattern: '^([1-9]|1[0-7])$' to prevent invalid input.
• Implement pagination for list_* tools: add 'limit' (default 50, max 500) and 'offset' (default 0) parameters. Return { data: [...], total: N, hasMore: bool }.
• Clarify discovery workflow in get_series_data description: 'First call list_indicators or list_series to find a seriesCode, then call list_geoareas to find an areaCode, then use this tool to fetch observations.'