Three tools with minimal documentation. All have descriptions (10-20 chars baseline), but lack depth and context. Input schemas present but sparse: 'search' and 'summarize' have basic parameter definitions; 'search_with_auth' has only a query string. No output schemas documented. No error handling guidance, no parameter constraints (enums, ranges), no idempotency hints. Tool names are action-verb based (search, summarize) which is good, but descriptions are too terse to guide LLM selection. The 'search_with_auth' description is domain-specific (authgent token chaining) but lacks actionable guidance on when to use it vs 'search'. No evidence of pagination, result limits, or composition patterns.

• Expand tool descriptions to 50-200 chars. For 'search': 'Search the corpus for documents matching a query. Returns up to max_results ranked by relevance. Use this for general discovery; use search_with_auth for protected resources requiring delegation tokens.'
• Document output schemas for each tool. E.g., search returns {results: [{id, title, snippet, score}], total_count, next_cursor}. Summarize returns {summary: string, word_count, confidence}.
• Add constraints to 'max_results': min=1, max=100, default=5. Document in description: 'Maximum number of results (1-100, default 5).'
• Add error handling guidance: 'If no results found, try broadening the query or check corpus availability. If summarize fails, ensure text is valid UTF-8 and under 100KB.'
• Clarify 'search_with_auth' use case: 'Use this variant when the downstream resource requires an authgent delegation token (RFC 8693). The token's act chain will include this MCP server, enabling full audit trails. For public resources, use search().'
• Consider adding pagination support (limit, offset/cursor) to search if results can exceed 100 items.

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

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