Four tools with clear verb-noun naming (search, get_entry, list_by_category, list_news) and solid descriptions (120-180 chars each). All tools have input schemas with typed parameters and descriptions. However, output schemas are completely undocumented, tools return raw API responses as strings with no structure guidance for LLMs. Parameter descriptions lack constraints (e.g., 'kind' accepts 'tool|skill|mcp' but no enum; 'limit' unbounded). No error handling guidance. Missing idempotency/destructiveness annotations. Baseline: 194 chars avg description (these are 120-180, acceptable); 100% of A+ tools document output schemas (this server: 0%).

• Document output schema for each tool. E.g., search returns {results: [{name, tagline, pricing, installSnippet}], total}. Structured output lets LLMs chain calls and extract data reliably.
• Convert 'kind' parameter to enum: kind: enum['tool', 'skill', 'mcp']. Add constraint descriptions to 'limit' (1-100, default 10/12).
• Add error handling: catch API errors and return actionable messages. E.g., 'Invalid kind: got "tools", must be one of: tool, skill, mcp. Try again with correct kind.'
• Add tool annotations via rmcp metadata. Mark all four tools with readOnlyHint=true and idempotentHint=true so LLM knows they are safe to retry.
• Add pagination guidance to list_by_category and list_news descriptions: 'Returns up to N items. Use limit parameter to control page size. Call multiple times with offset if needed.'

Inferred effective spec: 2026-07-28+.

• Official extensions adopted: Apps