Server has 8 tools with basic schemas and descriptions, but significant gaps limit production readiness. All tools have descriptions (10-50 chars, mostly under the 50-200 char LLM-optimized range), and input schemas are present with type definitions. However, output schemas are completely undocumented, LLMs cannot plan downstream calls or extract structured data. Error handling returns generic strings without recovery guidance. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Parameter descriptions are minimal (e.g., 'Search query' lacks format/length constraints). No pagination support for list-like tools (search_twitter, get_user_tweets, get_timeline). Rate limiting is implemented server-side but not exposed to the LLM, risking silent failures.

• Add output schemas to all tools. Document return type (e.g., 'Returns array of Tweet objects with fields: id, text, author, created_at, engagement_metrics'). Enable LLM to extract tweet_id for chaining into post_tweet(reply_to=...).
• Implement tool annotations: add readOnlyHint=true to search_twitter, get_user_tweets, get_timeline, get_latest_timeline; add destructiveHint=true to delete_tweet, delete_dm; add idempotentHint=true to post_tweet (if idempotent via deduplication key).
• Replace generic error strings with actionable recovery guidance. E.g., 'Failed to search tweets: Rate limit exceeded. Wait 15 minutes before retrying.' or 'User not found. Try search_users() with a partial name first.'
• Add pagination to list tools: accept offset/limit or cursor parameters; return total_count and next_cursor in response. Cap default count at 20-50 to prevent context window exhaustion.
• Enhance parameter descriptions with constraints. E.g., 'query (string, 1-500 chars): Search terms. Supports operators: from:@user, lang:en, -filter_word.' Use enum for sort_by: ['Top', 'Latest']. Add min/max for count: 1-100.

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

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