Two tools with clear verb-noun naming (search, user_tweets) and good descriptions (194 - 250 chars, within baseline 34 - 392). Input schemas present with types and defaults. However, output schemas are undocumented, LLMs cannot plan downstream calls or extract fields reliably. Error handling is present but generic; _wrap_twikit_error provides context but lacks actionable recovery steps. No parameter validation constraints (e.g., limit bounds). Descriptions mention X syntax support but lack examples of when to use each tool vs. the other.

• Document output schema in tool descriptions and/or via a separate schema definition. Example: 'Returns list of tweets with fields: id (string), text (string), author (string), created_at (ISO 8601), url (string), reply_count (int), retweet_count (int), favorite_count (int), view_count (int|null).'
• Add parameter constraints: limit should be integer, min=1, max=100 (or appropriate cap). Document in description: 'Maximum number of posts to return (1 - 100, default 20).'
• Enhance error messages with actionable next steps. Instead of returning raw exception, catch and wrap with guidance: 'Cookie auth failed. Run: xee-mcp init --force. See docs/cookies.md for details.' This text should be in the RuntimeError message, not just in a module-level constant.
• Add tool-level context in descriptions. For search: 'Use for keyword-based discovery across X. Supports advanced syntax (from:, since:, etc.). For a specific user's timeline, use user_tweets instead.' For user_tweets: 'Use to fetch recent posts from a known X handle. For keyword search across all users, use search instead.'
• Consider adding a third tool (e.g., get_user_info) to fetch user metadata (follower count, bio, verified status) without requiring a timeline fetch. This enables richer agent reasoning and reduces unnecessary data transfer.

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

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