Three tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. All tools have descriptions (10-20 chars baseline met), but descriptions lack actionable context for LLM selection. Input schemas present but incomplete: 'limit' parameter has a default but no min/max bounds. Output schemas are inferred from code (JSON dumps) rather than explicitly documented. No error handling guidance, no parameter validation rules stated, no recovery paths. Naming is verb-forward (reindex, search_code, index_status) but descriptions don't explain WHEN to use each tool or dependencies between them.

• Document output schemas explicitly. For search_code, state: 'Returns array of {file_path: string, start_line: int, end_line: int, content: string, chunk_type: string, name: string, score: float}'. For reindex, state: 'Returns {file_count: int, chunk_count: int, duration_seconds: float}'.
• Add bounds to 'limit' parameter: 'Maximum number of results (1-100, default 10)'. Prevents LLM from requesting 10000 results.
• Enhance tool descriptions with WHEN-to-use context. E.g., reindex: 'Call this when source code changes significantly or after first setup. Required before search_code can find results.' search_code: 'Call after reindex completes. Finds code by semantic meaning, not keywords.'
• Add error recovery hints to descriptions. E.g., search_code: 'If no results, try: (1) verify index exists via index_status, (2) reindex if stale, (3) simplify query.' reindex: 'If fails, check project_path exists and is readable.'
• Clarify project_path format: 'Absolute or relative path to project root. Must be readable and contain source files.' Add validation: reject paths outside allowed directories to prevent traversal attacks.