21 tools with inconsistent quality. All tools have descriptions (positive), but many are generic, translated from Chinese, and lack actionable guidance. Input schemas are present and properly typed for most tools, but descriptions are often under 100 chars and fail to explain WHEN to use each tool or what distinguishes similar operations. No output schemas documented. Error handling is absent, tools return raw API responses without recovery guidance. Parameter descriptions are minimal (e.g., 'container_name（字符串）：容器名称' is just a type restatement). No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Composition is reasonable (21 focused tools), but naming could be clearer (e.g., 'get_more_session_output' is vague about what 'more' means).

• Add output schemas to all tools. Document field names, types, and structure. Example: 'Returns: {"status": string, "container_id": string, "logs": string}'. This enables LLMs to chain tools and extract required data.
• Rewrite descriptions to be LLM-optimized (50 - 200 chars). Include WHAT the tool does, WHEN to use it, and any prerequisites. Example: 'Fetch logs from a running container. Use after create_container to debug startup issues. Returns raw stdout/stderr.'
• Add tool annotations to the FastMCP decorator. Use readOnlyHint=true for list_*, fetch_*, get_* tools; destructiveHint=true for remove_*, delete_*; idempotentHint=true for start_*, stop_*, restart_*. This guides LLM retry logic.
• Implement error handling with recovery guidance. Catch API errors and return structured responses like '{"error": "Container not found", "suggestion": "Call list_containers() to see available containers"}'. Categorize as retryable or user-fixable.
• Clarify ambiguous parameter names. Rename 'get_more_session_output' to 'fetch_session_output_chunk' or 'poll_session_output'. Replace 'other_files' with 'additional_files_json' and document the expected JSON structure in the description.