Cloison Runtime exposes 5 tools with significant quality gaps. Tool definitions are duplicated (memory_search and memory_store each appear twice), indicating registration issues. Schemas are present but minimal, parameters lack descriptions (query, maxResults, content, skillId, params all undescribed in schema). Descriptions exist but are generic and do not explain WHEN to use each tool or what downstream actions they enable. The skill_execute tool accepts an untyped 'params' object, violating constrained-input patterns. No error handling guidance, no pagination support, no output schema documentation. This is a D-grade server: definitions are skeletal, not LLM-optimized.

• Remove duplicate tool registrations. Ensure memory_search and memory_store are registered exactly once in createMemoryTools().
• Add descriptions to all schema properties: query ('Search query string (max 10,000 chars)'), maxResults ('Max results to return, 1-100, default 5'), content ('Content to store, max 1,000,000 chars'), skillId ('Skill identifier to execute'), params ('Skill parameters as JSON object').
• Document output schemas in tool descriptions. E.g., memory_search: 'Returns array of {id, content, relevance_score}'; memory_store: 'Returns {id: string}'.
• Replace skill_execute's untyped 'params' with explicit properties or an enum of known skill IDs with their parameter schemas. If dynamic, document the expected structure in the description.
• Enhance error messages with recovery guidance. E.g., 'Query too long (10,500 chars, max 10,000). Try a shorter query or use maxResults to limit results.' instead of bare error strings.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry