Server has 8 tools with basic schemas and descriptions in Chinese. All tools have input schemas with type definitions and descriptions present. However, descriptions are minimal (10-30 chars), parameter descriptions lack actionable detail, and output schemas are not documented. No tool annotations (readOnlyHint/destructiveHint). Error handling exists but lacks recovery guidance. Tool naming follows verb_noun pattern (start_, get_, cancel_) which is good, but descriptions are too terse to guide LLM selection effectively. No evidence of idempotency guarantees or composition patterns.

• Expand all tool descriptions to 80-150 chars. Example: 'start_architecture_analysis' → 'Submit a list of application IDs to begin asynchronous architecture analysis. Returns a task_id to track progress. Analysis includes dependency mapping, component analysis, and report generation.'
• Document output schemas inline or in a separate schema registry. For get_task_status, specify: returns {task_id, status, progress, stage, message, created_at, updated_at}. For get_task_result, specify: returns {task_id, success, result, error_message}.
• Add tool annotations: mark start_architecture_analysis and cancel_task with destructiveHint=true; mark all get_* tools with readOnlyHint=true. This enables LLMs to reason about retry safety.
• Enhance error messages with recovery paths. In get_task_status 404 handler, return: 'Task "<task_id>" not found. Call get_tasks() to list active tasks, or check the task_id spelling.'
• Add parameter descriptions for task_id (e.g., 'UUID returned by start_architecture_analysis') and report_format (e.g., 'markdown: human-readable report; json: structured data for parsing; html: web-viewable report').

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

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