23 tools with complete schemas and descriptions, but significant quality gaps. All tools have input schemas with types and descriptions, meeting baseline requirements. However, descriptions are often generic or translated (Korean→English), lacking LLM-optimized clarity. Tool naming follows verb_noun convention (get_, search_, send_) but lacks distinction between similar operations (e.g., health_check, get_stats, get_config appear 3x across services). No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. Error handling and recovery guidance absent. Output schemas undocumented. Parameter constraints present (enums, min/max) but descriptions don't always explain format/range clearly.

• Add tool annotations to all tools: readOnlyHint=true for all READ_ONLY tools; destructiveHint=true for slack_send_message; idempotentHint=true for all GET operations.
• Consolidate or rename duplicate utility tools. Either create single service-agnostic health_check/get_stats/get_config tools, or rename to confluence_health_check, jira_health_check, slack_health_check (already done for some).
• Expand descriptions to 80-150 chars with context: 'Search Confluence pages by CQL query. Use when you need to find content by title, author, or date. Returns paginated results with page ID, title, and URL for chaining to confluence_get_page.'
• Document output schemas in tool descriptions or separate schema docs. E.g., 'Returns {results: [{id, title, url, space_key}], total_count, cursor}' so LLMs know what fields to extract.
• Add error handling guidance to descriptions: 'If search returns 0 results, try broader keywords or check space permissions. If auth fails (401), verify Confluence API token is valid.'

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

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