Monorepo contains 14 tools across three servers. Most tools have descriptions (13/14), but descriptions are generic and lack actionable context for LLM selection. Input schemas are present but lack depth: many parameters missing descriptions, no enums for constrained inputs, no min/max bounds. Output schemas are not documented. Tools like 'sequentialthinking' and 'mental_model' have vague, non-action-oriented names. Error handling and recovery guidance absent. No tool annotations (readOnlyHint, destructiveHint). Composition is weak, tools do not chain well (e.g., no IDs returned to enable downstream calls). Security: no evidence of secret injection or permission gates.

• Rename abstract tools to action-verb form: 'sequentialthinking' → 'analyze_problem_sequentially', 'mental_model' → 'apply_mental_model', 'metacognitive_monitoring' → 'assess_reasoning_quality'. This enables LLMs to infer intent from the name alone.
• Add comprehensive parameter descriptions for all array/object types. For 'personas' in collaborative_reasoning, specify: 'Array of objects with keys: id (string), name (string), expertise (string), perspective (string)'. For 'elements' in visual_reasoning, document node/edge structure.
• Document output schemas for all tools. Example for mental_model: 'Returns {analysis: string, steps: string[], conclusion: string, confidence: number (0-1)}'. This enables LLMs to extract and chain results.
• Convert description-based enums to JSON Schema enums. Move 'first_principles, opportunity_cost, ...' from mental_model description into a formal enum array. Yelp tools already do this correctly, apply the same pattern.
• Add error handling with recovery guidance. Example: 'If analysis fails, return {error: string, suggestion: string, retryable: boolean}'. For Yelp tools, add: 'If business_id not found, return available alternatives or suggest search_yelp_business first.'