Server has 9 tools with mostly complete schemas and descriptions in Chinese. Tool names follow verb_noun pattern (query_*, update, status). Descriptions are detailed (100-300 chars) but written for Chinese users, not LLM optimization. All tools have input schemas with types. Critical gaps: no output schemas documented, no error recovery guidance, no parameter enums for constrained inputs (e.g., query_type accepts free-form strings instead of enum), and no tool annotations (readOnlyHint/destructiveHint). Parameter descriptions are present but lack format/constraint details. Two tools (arknights_data_update, arknights_query_alias) modify state but lack confirmation/dry-run patterns.

• Add JSON Schema definitions for all tool outputs. Document response structure: {ok: boolean, message: string, data: <type>, card?: {url: string, ...}}. Specify data field types for each tool (e.g., operator info returns {name, rarity, profession, skills: [...], ...}).
• Convert query_type parameter to enum: ['info', 'cost', 'skills', 'tokens', 'module', 'skin']. Remove free-form string acceptance. Update schema to use 'enum' constraint.
• Add tool annotations: mark arknights_data_status and all query_* tools with readOnlyHint=true; mark arknights_data_update and arknights_query_alias with destructiveHint=true.
• Enhance error messages with recovery guidance. E.g., 'Operator not found. Try arknights_query_alias list operator to see registered aliases, or search with a partial name.' Include available alternatives when resources not found.
• Add dry-run support to arknights_data_update (wait=false, dry_run=true returns what would be updated without committing) and arknights_query_alias register (returns confirmation prompt before persisting).

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

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