Server has 20 tools with complete input schemas and descriptions in Chinese. Naming follows verb_noun pattern (check_, get_, publish_, post_, etc.). However, descriptions lack LLM-optimized depth, most are 1-2 sentences without WHEN/WHY context or prerequisites. Parameter descriptions are present but generic (e.g., 'Feed ID' without format hints). Output schemas are not documented in the source. Error handling guidance is absent. Tool composition has issues: publish_content and publish_video are separate but nearly identical; broadcast_publish duplicates publish_content logic. No pagination support visible for list_feeds or search_feeds despite potentially large result sets.

• Document output schemas for all tools. For list_feeds, specify: {feeds: [{feed_id, title, author_id, xsec_token, ...}], total_count, next_cursor}. This enables LLMs to chain calls (e.g., list_feeds → get_feed_detail).
• Expand descriptions to 50-200 chars with WHEN/WHY context. E.g., 'get_feed_detail: Fetch full content, comments, and engagement metrics for a feed. Call after list_feeds to get details. Requires feed_id and xsec_token from list_feeds response.'
• Add pagination to list_feeds and search_feeds: limit (1-100, default 20), offset, and total_count in response. Cap results at 50 items per call to prevent context window exhaustion.
• Unify publish_content and publish_video into a single publish_note tool with a 'media_type' enum (image|video) and 'media' param (array for images, single for video). Or clearly document why they must be separate.
• Add error handling guidance to destructive/write tools. E.g., publish_content: 'Returns error if images are invalid or title exceeds 20 chars. Retry with corrected inputs. If schedule_at is in the past, use current time instead.'

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

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