Single tool 'codebase_search' has a complete JSON Schema with all parameters typed and described in Chinese. Naming follows verb_noun pattern (search_*). Description is detailed (~280 chars) but in Chinese, limiting LLM comprehension in English-first contexts. Schema includes proper types (string, integer, boolean), defaults, and descriptions. However, output schema is not documented, LLMs cannot predict return structure. No error handling guidance. No tool annotations (readOnlyHint, etc.). Parameter descriptions are functional but lack constraint examples (e.g., glob pattern syntax, score range). Missing English descriptions is a moderate gap for production use.

• Add English translations for tool and parameter descriptions. Provide both Chinese and English to maximize LLM compatibility.
• Document the output schema explicitly. Specify return fields (e.g., results: [{file_path, score, code, start_line, end_line, symbol, language, related_calls}], total_count, next_cursor). Include field types and descriptions.
• Add error handling guidance to the tool description: 'Returns error if directory_path does not exist or is not readable. If embedding service fails, returns partial results with available chunks. Retry with smaller top_n if timeout occurs.'
• Add tool annotations: mark as readOnlyHint=true (search does not modify state) and idempotentHint=true (same query always returns same results).
• Specify parameter constraints: top_n range (1 - 100, default 10), path_filter examples ('*.py', '**/service/**', 'controller'), and clarify that include_related defaults to true for call graph expansion.

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

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