Three tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. memory_recall and memory_store have reasonable descriptions (40-60 chars), but memory_forget lacks clarity on destructive behavior. All three tools have input schemas with types and enums, but parameter descriptions are minimal (10-20 chars). No documented output schemas. Error handling and recovery guidance absent. Naming follows verb_noun pattern (positive), but descriptions do not explain WHEN to use each tool or what data is returned.

• Add explicit output schema documentation for each tool. Example for memory_recall: 'Returns array of {id: string, content: string, score: number (0-1), timestamp: ISO8601}. Limit: 20 results max.'
• Expand memory_forget description to: 'Permanently delete memories by URI or semantic query. WARNING: This operation is irreversible. Provide either uri (exact match) OR query (semantic search), not both. Returns count of deleted memories.'
• Increase parameter descriptions to 50-100 chars with format hints. E.g., 'query: semantic search string (1-500 chars, e.g., "user preferences for coffee"). Matches against memory content and tags.'
• Document memory_store return value: 'Returns {id: string, stored_at: ISO8601, scope: string}. Use id in future memory_recall or memory_forget calls.'
• Add error guidance: 'If memory_recall returns empty, try broader query terms or check scope (user vs agent). If memory_forget fails with "not found", use memory_recall first to verify memory exists.'

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

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