Citable MCP has 9 tools with basic schemas and descriptions, but significant quality gaps. All tools have descriptions (10-80 chars, mostly short), and input schemas are present with type definitions. However, descriptions are uniformly terse and lack actionable context for LLM selection. No output schemas are documented. Parameter descriptions are minimal (e.g., 'The domain to check rankings for' provides no format guidance). Error handling is absent from tool definitions. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all being read-only. Tool naming follows verb_noun convention (citable_*) but is domain-specific and could be clearer for cross-domain agents. The citable_connect tool (sign-in flow) lacks clarity on what 'reset' does and when to use it. No pagination, batch operations, or composition patterns evident.

• Document output schemas for all 9 tools. Include field names, types, and descriptions. E.g., citable_rankings should return {domain, keywords: [{keyword, rank, volume}], traffic_estimate}. This enables LLMs to chain tools and extract the right data.
• Expand tool descriptions to 100-200 chars. Add WHEN to use (e.g., 'Use this to discover keyword opportunities before creating content'), WHAT it returns (e.g., 'Returns top 20 keywords with monthly search volume'), and any prerequisites (e.g., 'Requires a funded wallet').
• Add readOnlyHint: true to all 9 tools in the tool definition. This signals to agents that these tools are safe to retry and compose without side effects.
• Enhance parameter descriptions with format guidance. E.g., 'domain: The domain to check (e.g., example.com, without https:// or www)' and 'topic: The search topic (2-50 characters, e.g., "machine learning frameworks")'.
• Add error recovery guidance to tool descriptions. E.g., 'If the domain is not found, try a different domain or check citable_prices to see if the domain is supported.' This helps agents self-correct.