13 tools with minimal descriptions (all ~20-30 chars, below baseline 194 chars) and severely underdeveloped schemas. All tools accept only a single 'query' string parameter with no enums, ranges, or validation constraints. No output schemas documented. Descriptions lack WHEN/WHY context, prerequisites, or recovery guidance. No error handling patterns visible. Tool names use 'ris_' prefix (domain-specific) rather than action verbs (search_, get_), reducing clarity for LLM selection. No tool annotations (readOnlyHint, etc.). Composition is repetitive: 11 of 13 tools are search variants with identical structure, violating single-responsibility principle.

• Rename tools to action verbs: search_austrian_federal_law, search_austrian_state_law, search_case_law, etc. This enables LLMs to infer intent from the name alone.
• Expand all descriptions to 100 - 200 characters. Include: (1) What does it search? (2) When should I use it instead of similar tools? (3) What does it return? Example: 'Search Austrian federal laws (Bundesrecht) by keyword. Use this for national legislation; use ris_landesrecht for state laws. Returns matching documents with title, date, and document number.'
• Document output schemas for all tools. Specify: result structure (array of objects), per-item fields (title, date, dokumentnummer, url), pagination (limit, offset, total_count), and result limits (e.g., max 50 results per call).
• Add pagination parameters (limit, offset) and result limits to search tools. Cap results at 20 - 50 items per call. Document in description: 'Returns up to 50 matching documents. Use offset for pagination.'
• Add tool annotations to schema: set readOnlyHint=true for all tools since they only read from RIS API. This signals to agents that these calls are safe to retry.