Single tool 'convert' has a detailed description (Japanese, ~500 chars) and comprehensive input schema with 7 parameters, all typed and described. However, critical gaps prevent higher scoring: (1) Tool name 'convert' is generic and does not follow verb_noun pattern (should be 'convert_file' or similar); (2) Output schema is completely undocumented, no specification of what the tool returns on success or failure; (3) Parameter descriptions are in Japanese, creating accessibility barriers for English-speaking LLM agents; (4) No error handling guidance, tool description mentions 'stops' on conflicts but does not explain recovery paths; (5) No indication of idempotency, retry safety, or side effects; (6) Missing parameter constraints like file size limits or timeout behavior. Schema is present and mostly well-formed (types, descriptions, min/max on numeric fields), but output documentation is absent, which violates the 'Document the output schema' critical check.

• Rename tool from 'convert' to 'convert_file' to follow verb_noun pattern and improve LLM intent recognition.
• Add complete output schema documentation: specify return type (object with fields like 'output_path', 'format', 'size', 'duration' as applicable), success indicators, and error response structure.
• Translate all parameter descriptions from Japanese to English. Provide both languages if needed, but English must be primary for agent compatibility.
• Add explicit error handling guidance: document error codes/messages for common failures (unsupported format, file not found, disk full, permission denied) and recovery steps for each.
• Document idempotency and retry safety: clarify whether calling convert twice with identical inputs produces one output or two, and whether the tool is safe to retry on transient failures.