Server defines 5 tools with complete input schemas and descriptions. Naming follows verb_noun convention (list_, describe_, search_, retrieve_, install_). Descriptions are present but brief (10-60 chars), below the 50-200 char LLM-optimized baseline. All parameters have type definitions and descriptions. However, output schemas are not documented, LLMs cannot predict response structure. Error handling exists but lacks recovery guidance. install_shared_skill combines multiple concerns (fetch + install + audit) and accepts an overwrite flag defaulting to true, risking unintended data loss.

• Document output schemas for all 5 tools. Example: 'Returns {skills: [{name, category, description, version}], total_count, next_cursor}' so LLMs know what fields to extract.
• Expand tool descriptions to 50-200 chars. For list_shared_skills: 'List available skills in the registry, optionally filtered by category. Returns paginated results with skill metadata. Call this first to discover available skills before retrieving or installing.'
• Split install_shared_skill into two tools: fetch_shared_skill_bundle (read-only, returns bundle metadata) and confirm_install_shared_skill (requires explicit confirmation, performs install). Prevents accidental overwrites.
• Add recovery guidance to error messages. Instead of 'install was not authorized', return 'Install not authorized by registry. Check SSR_MCP_AUTH_TOKEN or contact registry admin. Retry with valid credentials.'
• Document pagination behavior: 'Returns up to limit results. If total_count > limit, use offset parameter in next call to fetch remaining results.' Add offset parameter to list_shared_skills and search_shared_skills.